51.Rehabilitation Engineering

51• Rehabilitation Engineering 51• Rehabilitation Engineering Artificial Hearts and Other Organs Abstract | Full Text: ...

Author: John G. Webster (Editor)

58 downloads 1561 Views 2MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!
Report copyright / DMCA form

DOWNLOAD PDF

51• Rehabilitation Engineering

51• Rehabilitation Engineering Artificial Hearts and Other Organs Abstract | Full Text: PDF (235K) Artificial Limbs Abstract | Full Text: PDF (325K) Assistive Devices For Motor Disabilities Abstract | Full Text: PDF (403K) Hearing Aids Abstract | Full Text: PDF (236K) Human Motion Analysis Abstract | Full Text: PDF (652K) Prosthetic Power Supplies Abstract | Full Text: PDF (285K)

file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20ELEC...S%20ENGINEERING/51.Rehabilitation%20Engineering.htm15.06.2008 19:54:34

file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20ELECTRICAL%...20ENGINEERING/51.%20Rehabilitation%20Engineering/W6602.htm

}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP

Home / Engineering / Electrical and Electronics Engineering

Wiley Encyclopedia of Electrical and Electronics Engineering Artificial Hearts and Other Organs Standard Article Roger P. Gaumond1, John F. Gardner2, Alan J. Snyder3, William J. Weiss4, Christopher Kelley5, Dennis R. Trumble5 1Penn State University, University Park, PA 2Penn State University, University Park, PA 3Penn State University, Hershey, PA 4Boston University School of Medicine, Boston, MA 5Allegheny University of the Health Sciences (Allegheny Campus), Pittsburgh, PA Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6602 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (235K)

●

●

● ●

Recommend to Your Librarian Save title to My Profile Email this page Print this page

Browse this title ●

Search this title Enter words or phrases

Abstract The sections in this article are Energetics of Blood Pumping Control of Cardiovascular Pressure and Flow Approaches to Mechanical Design of Artificial Hearts Simulation of Pump Performance and TAH Control Ventricular Assist Energy Utilization Transcutaneous Energy Transmission and Batteries Cardiomyoplasty Fundamentals and Muscle-Powered Blood Pump

file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20EL...RING/51.%20Rehabilitation%20Engineering/W6602.htm (1 of 2)15.06.2008 19:55:46

❍

❍ ❍

Advanced Product Search Search All Content Acronym Finder


Other Artificial Organs About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.


656

ARTIFICIAL HEARTS AND OTHER ORGANS

this technology is severely limited by the shortage of suitable donor organs. Artificial organ systems, constructed of humanmade materials, are generally not subject to rejection problems, nor is organ availability of prime concern. However, their development and use poses its own set of challenges. The heart, liver, kidney, and other organs perform functions which are not fully understood by modern science. Artificial organs for implant therefore provide at best an incomplete solution to a medical problem, but a solution which can be satisfactory in some cases. Such intervention generally involves the implantation of nonbiological elements within the body, raising issues of biocompatability, and requiring stringent sterilization protocols. The use of implanted humanmade structures, such as heart valves in contact with circulating blood, has been associated with an increased incidence of thromboembolic events. This article considers the energy requirements, and the sensor and control aspects of the design of artificial hearts and of systems designed to augment the cardiovascular pumping provided by a failing heart. Sensor, control, and energy requirements are common to other organs as well, and these requirements will be discussed in relation to the artificial pancreas and other artificial organs. ENERGETICS OF BLOOD PUMPING The heart does a substantial amount of work in pumping blood through the body. Pumping a mean flow, or cardiac output (CO), of 5 L/min to an aortic pressure head of 13.3 kPa (100 mm Hg) requires a work rate of 1.1 W. The most energydense batteries available, lithium primary cells, have a capacity less than 500 Wh/kg. Nearly 20 kg of batteries would be required to store the energy for one year of operation under the very best of conditions. In fact, neither the natural heart nor the human-made pumps designed to replace it are particularly energy efficient. Obtainable efficiencies vary with CO, ranging from about 10 to about 25% for natural and human-made analogs. Therefore, a substantial energy source must be provided for blood pumping. Transcutaneous energy transmission using magnetic induction coupling is one means of powering an implanted pump and is the preferred system for proposed human-made blood pump implants. The energy derived from metabolism of food is another, and is the source of energy for proposed muscle-powered blood pumps. CONTROL OF CARDIOVASCULAR PRESSURE AND FLOW

ARTIFICIAL HEARTS AND OTHER ORGANS Organ transplantation from suitable donors is a medical technology posing risks of tissue rejection which are well-handled today by suitable drug regimen. Widespread application of

The mammalian cardiovascular (CV) system has evolved elaborate and redundant systems for dealing with shifts in physiological need. When demand for gas exchange increases, the CO may increase from 5 to 15 L/min in response. The force of contraction is related to the degree of ventricular filling (preload) through the Frank–Starling mechanism. In addition both the force of contraction and the heart rate are modified appropriately by the rapid response of the autonomic nervous system, and by a slower but longer acting endocrine response. In a healthy CV system, this CO increase is carried out while maintaining the blood pressure within a narrow range. For this to occur, the CO and the hydraulic impedance of the cardiovascular system change in opposite directions.

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.


Heart R Cao

Ca

Figure 1. Heart pumping into a 3-element lumped-parameter circulation model representing aortic Cao and atrial Ca compliances and systemic resistance R.

Consider the simplified one-sided lumped-parameter CV system consisting of aortic and venous compliances and a systemic resistance in Fig. 1. The venous compliance includes the right atrium, and has been labeled Ca. The pulmonary circulation is included as part of the heart pump in this representation. The circulating blood volume (BV) consists of that within the aortic compliance (PaoCao) plus that within the venous compliance (PaCa) plus the mean volume in the heart, BH. The following relations apply, where all quantities are to be interpreted as their mean values: BV = BH + PaoCao + PaCa

(1)

Pao − Pa = CO × R

(2)

Equations (1) and (2) combine to yield: Pao + 1/Ca (BH − BV) = (Ca /(Cao + Ca ))(CO × R)

657

chamber designs have been developed (Table 1). Early designs using large pneumatic drivers to provide alternating pressure and vacuum pulses to a drive line (flexible tubing) attached to the pump housing are in common use, but patient mobility is limited. Brushless dc motors (BDCM) or solenoid actuators coupled directly to the pumping chamber allow patients sufficient mobility to return to normal activities. In the newer systems currently under development (1–4), power is inductively transferred to an implanted controller, thereby eliminating the risk of infection associated with percutaneous cable sites. Continuous flow pumps use high-speed rotating impellers in the blood stream to produce axial or centrifugal flow without the need for valves. These pumps have the potential to be significantly smaller than positive displacement pumps. Developers of systems intended for long-term implantation include: Nimbus (Rancho Cordova, CA) and the University of Pittsburgh (5); Transicoil, Inc. (Valley Forge, PA) and the Texas Heart Institute (Houston) (6); the Cleveland Clinic Foundation and the University of Utah (7); Baylor College of Medicine and NASA/Johnson Space Center (Houston) (8). Current research focuses on reducing fluid stresses on the blood, developing durable and antithrombogenic bearings, and refining control methods. One design approaching in vivo durability testing in total artificial heart (TAH) and ventricular assist device (VAD) configurations is the rollerscrew/pusherplate system developed by Penn State/3M Healthcare/Arrow International and about which we will focus our examples of calculations. Placement of motor, electronics/battery cannister, transcutaneous transmitter, and an external case for carrying batteries is illustrated in Fig. 2 for the TAH System.

(3)

The natural cardiovascular control system employs a distributed network of baro- and chemoreceptors to respond to increased oxygen need by increasing heart rate and cardiac contractility which may increase CO by a factor of three within a few seconds. There is a need to maintain Pao within narrow limits, and blood volumes do not change rapidly. The CV control system responds to CO increases by decreasing both R and Ca so as to keep Pao at a safe level as shown in Eq. (3). Mechanical systems meant to augment or replace the heart are faced with the dual challenges of responding to changes in flow requirements and providing a means for pressure to remain within safe limits despite these flow changes.

APPROACHES TO MECHANICAL DESIGN OF ARTIFICIAL HEARTS Circulatory support devices encompass a wide variety of pump designs, which reflect various approaches to the method of actuation, anatomical placement, and desired blood flow patterns within the pumping chamber. Devices may be generally categorized as either positive displacement or continuous flow. Positive displacement pumps produce pulsatile flow in similar fashion to the natural heart, with alternate filling and ejection phases. Inlet and outlet valves maintain unidirectional blood flow. A variety of actuation methods and pump

SIMULATION OF PUMP PERFORMANCE AND TAH CONTROL Human-made blood pumps are often performance-tested using a mock circulatory loop, a group of mechanical elements which simulate the hydraulic impedance of the cardio-vascular system. An alternate approach is the use of computer simulation. This approach has the advantage of allowing hydraulic compliance and resistance elements to change rapidly to values which reflect cardiovascular control response. We show here some examples of such calculations. The narrative in the section on TAH control was copyrighted in 1995 by IEEE. Reprinted, with permission, from proceedings of the NE Bioengineering Conference, Bar Harbor, ME, May, 1995, pp. 22–23. Physically, the TAH we consider consists of a small brushless dc motor driving an efficient rotary to linear conversion mechanism. The resulting motion drives a circular pusherplate which compresses a flexible ventricle against a rigid case. Pusherplates are attached to both sides of the motor drive, so that blood is ejected from one ventricle as the other is passively filling. The controller of this TAH has three main tasks: controlling the applied motor voltage on a stroke to stroke basis so that the pusherplate follows a desired trajectory, maintaining balance between the systemic and pulmonary circulations by manipulating the relative filling periods of the left and right ventricles, and varying overall cardiac output in response to demand (9,10). Studies have shown that, due to elevated venous pressures, an output control method based on the Star-

658


Table 1. Circulatory Support Systems Utilizing Positive Displacement Pumps Currently under Advanced Development for Bridge-to-Transplant and/or Long-Term Use Developer

Type

Thermo Cardiosystems (Woburn, MA) Thermo Cardiosystems (Woburn, MA) Thoratec Laboratories Corp. (Berkeley, CA) Novacor Division, Baxter Healthcare (Oakland, CA) CardioWest Technologies (Tucson, AZ) Penn State University (Hershey and University Park, PA); 3M HealthCare (Ann Arbor, MI); Arrow International (Reading, PA) Abiomed (Danvers, MA); Texas Heart Institute (Houston, TX) Nimbus (Rancho Cordova, CA); Cleveland Clinic (Cleveland, OH)

Force Transducer

Coupling to Blood Pump

VAD VAD VAD

Low-speed unidirectional BDCM External pneumatic driver External pneumatic driver

Helical cam 씮 pusherplate 씮 diaphragm Drive line 씮 pusherplate 씮 diaphragm Drive line 씮 diaphragm 씮 blood sac

VAD

Pivoting solenoid

TAH TAH and VAD

External pneumatic driver Moderate-speed bidirectional BDCM

Spring arms 씮 dual symmetric pusherplates compressing blood sac Drive line 씮 diaphragm Rollerscrew 씮 pusher plate(s) 씮 blood sac(s)

TAH

High-speed unidirectional BDCM

TAH

High-speed unidirectional BDCM

Axial flow pump 씮 hydraulic fluid 씮 spool valve 씮 blood sacs Gear pump 씮 hydraulic fluid 씮 magnetically coupled pusherplates 씮 diaphragms

Note: The first four systems have achieved a significant level of clinical use.

ling mechanism is insufficient for TAH recipients (10). Therefore, demand is sensed through changes in the systemic arterial pressure. Adjustments are made at intervals of one or more cardiac cycles, with the adjustment of pusherplate velocity (inner loop control) applied more frequently than left–right balance (preload) or demand–response (afterload) control. Direct measurements of pressures in the TAH and the circulatory system would require additional sensors and transducers, increasing the risk of device failure. It is therefore desirable to estimate these values from the applied voltage and current waveforms of the motor. This requires an accurate representation of motor parameters, and a reasonable representation for the CV system so that time-varying parameters can be inferred, and motor control adjusted.

Pra), and left and right ventricular pressures (Plv, Prv). The fluid-storing properties of the blood vessels are represented by capacitances (compliances), and resistances Rpulm and Rsys represent the pressure drop across the pulmonary and systemic capillaries. For the purposes of this model, valves are assumed to open and close instantaneously on pressure differentials. The pressure drop across an open outlet valve is represented by 3 elements in Fig. 3: P = Rov1Q + Rov2 Q2 + Lout

Control electronics and batteries

TAH motor

TETS coils

(4)

where Q is the flow across the valve, Rov1 and Rov2 are resistance terms, and L is the inertance of the fluid in the outlet graft. The pressure gradient across an open inlet valve, as well as for steady backflow across a closed valve is described as a square law relationship:

TAH Model A representation for the CV system driven by a TAH is shown in Fig. 3. Key pressures are aortic pressure (Pao), pulmonary arterial pressure (Ppa), left and right atrial pressures (Pla,

dQ dt

P = RQ2

(5)

Here, R varies according to whether the valve is open or closed, as well as with the diameter of the valve, which is different in the inlet and outlet positions. The ventricles are modeled as pressure-dependent compliances Clv(Plv) and Crv(Prv), with values determined heuristically in static tests. These tests demonstrate that as the pressure in the ventricle increases, the compliance decreases rapidly, as the flexible blood sac is compressed against the rigid case of the ventricle. The pusherplates act as a flow source (area ⫻ velocity) when coupled to the blood sacs, driving blood through the outlet ports. This analysis results in 80 sets of 12 differential equations, with the appropriate set selected depending on the position of the valves and the direction of pusherplate motion. The equations are readily solved by computer using a Runge–Kutta integration method. Pressure and flow waveforms from such a model compare quite favorably with measurements taken from mock circulatory loop tests (11–14).

Battery pack Figure 2. Elements of an implanted TAH system include an implanted blood pump, a means of transcutaneous energy transmission using magnetic induction, and an external battery.

TAH Control Simulation The velocity profile of the pusherplate is regulated by an adaptive feedforward control mechanism, where the applied


659

(Alp)v Rov1

Rov2

Lout

Riv2 Pao

Plv

Rsys

Lin

RovL

Cao

Clv (Plv) (–Arp)v

RivL Cra

Lout

Riv2 Ppa Rpulm

Lin

RovL Crv (Prv)

Pla

Cpa

Kb v Kc

Cla

(6)

where R is the winding resistance, i is the winding current, Kb is the motor’s back emf constant, and Kc is a constant relating the angular velocity of the motor with the linear velocity v of the pusherplate. The motor’s torque is related to the current by T = Kt i

(7)

where Kt is the motor’s torque constant. An estimate of the ventricular pressure is formulated by breaking the motor torque into frictional, speed, and load-dependent components, and assuming the force exerted by the load is equal to the pressure inside the ventricle multiplied by the area of the pusherplate. Solving for ventricular pressure yields

Kt R

K D J dv em − b v − − Tf v– Kc Kc Kc dt

Figure 3. Analog of the TAH coupled to the circulatory system. (From Ref. 15, courtesy of IEEE  1995 by IEEE. All rights reserved.)

RivL

motor voltage scheduled for the next stroke cycle is adjusted by a factor proportional to the difference between the desired and actual pusherplate velocities during the previous cycle. Control is scheduled at constant position intervals, corresponding to the commutations of the dc motor, rather than at fixed times. An estimate of the left ventricular pressure is obtained as follows. For a dc motor, the voltage applied to the motor is em = Ri +

To LV

(8)

where Alp is the area of the pusherplate, D is a damping coefficient, J is the rotary inertia of the moving parts of the motor, and Tf is a constant frictional torque. Equation (8) forms the basis of the output and balance control methods of the TAH (9,10,15,16). An example of the model’s computed ven-

tricular pressure signal is shown in the upper panel of Fig. 4. The lower panel shows the degradation in this signal caused by the limited sampling of position provided by the motor position sensors. As shown in Fig. 4, when the ventricle fills incompletely, the pusherplate attains considerable forward velocity before coming in contact with the ventricle. This event causes a large initial peak in the ventricular pressure, which damps out quickly. The pressure estimation deviates from that obtained from the model due to the limited rate of the asynchronous position sampling and the inaccuracies this introduces in velocity and acceleration estimation. Nevertheless, the initial pressure peak indicating contact with the pusherplate and blood sac is well represented. The control algorithm ad-

Pressure (mmHg)

Rov2

Prv

1 Pv(est) = Kc Alp

To RV

Rbron

Rov1 From RA

Pra

Pressure (mmHg)

From LA

300 200 100 0 0

0.05

0.1

0.15 0.2 Time (sec)

0.25

0.3

0

0.05

0.1

0.15 0.2 Time (sec)

0.25

0.3

300 200 100 0 –100

Figure 4. Systolic left ventricular pressure from TAH model (top) and estimation of degradation in pressure signal caused by limited sampling of position (bottom). (From Ref. 15, courtesy of IEEE  1995 by IEEE. All rights reserved.)

660


65

SV(ml)

60 55 50 45

0

5

10

15 20 25 30 Number of beats

35

40

Figure 5. Plot of stroke volume (SV) versus time. The pulmonary peripheral resistance is increased at beat 10. (From Ref. 15, courtesy of IEEE  1995 by IEEE. All rights reserved.)

justs the filling times as needed. If there is incomplete left ventricular filling, the filling times are altered to restore balance. In order to suppress the cardiac output’s dependence on right atrial pressure, the left pump filling time is increased when the right pump filling time is decreased, and vice versa. This balance control is implemented every five beats to allow averaging of these fill-rate indicating signals. This balance control adjustment requires a finite amount of time to implement. This can be shown in the model by allowing a step change in an element (Rpulm) of the model. The increase in resistance causes a subsequent reduction in left atrial pressure, which decreases ventricular filling, causing a left–right ventricular imbalance. Figure 5 shows a plot of left ventricular stroke volume versus the number of pump cycles when pulmonary resistance was doubled at beat 10. The decrease in filling pressure causes a steep reduction in the stroke volume. By increasing the left filling time while decreasing the right ventricular rate of filling, the balance control mechanism alone can correct the imbalance. At beat one, the left diastolic time is 0.36 s, but by beat 40 it has been increased to 0.44 s. There are many aspects of this model of TAH and cardiovascular interaction that need exploration including the dynamics of interaction between autonomic blood pressure regulation and the TAH system’s candidate control algorithms. VENTRICULAR ASSIST It is a relatively simple matter to apply the control techniques designed for the TAH to a ventricular assist device (VAD), which accepts blood from the left ventricular apex and returns it to the aorta. The assist device developed by the authors makes use of the same electric motor and force transducer used for the TAH, but has only a single pump and a single pusherplate on the electromechanical drive. Such devices are useful in the large population of patients whose ventricular failure is concentrated in the left ventricle. The control system for the ventricular assist device, like that for the TAH, must control the pusherplate motion trajectory (inner loop control) and the pump output (outer loop control). The inner loop control algorithm can be implemented as a feed-forward controller as with the TAH. The desired speed trajectories are somewhat different, since the device is exposed to different pressures. In systole, differences occur only due to the additional outflow tract resistance and inertance presented by the outlet cannula that takes blood from the ab-

dominally placed pump to the thoracic aorta. In VAD diastole, the motor is loaded only by the friction and inertia of the mechanism, and so the desired motion trajectory, particularly when energy efficiency is considered, may be substantially different from that used when a pulmonary load is expected. The outer loop control for the VAD is a subset of the set of algorithms used for the TAH. The primary goal of the VAD output control is to maintain a low pressure in the native ventricle. In doing so, the VAD (1) by definition, accepts all of the output provided by the right heart, and thereby prevents elevated pulmonary pressures, and (2) maintains low pulmonary venous pressures so that the right heart, which is typically afterload-sensitive, can provide as much cardiac output as possible. Outer loop control of the VAD is accomplished by monitoring left pump filling via Eq. (8), while making incremental increases and decreases in the pump rate. A decrease in pump filling results in an incremental decrease in the pump rate, with the expectation that filling of the pump will be restored. With the pump filling adequately, trial increases in rate are made in order to determine whether a higher pump output can be supported.

ENERGY UTILIZATION Kinetic energy imparted by the blood pump to the blood is delivered through a transduction chain that includes the inductive energy transmission system, the power electronics, mechanical force transducer, and finally the blood pump itself. The energy dissipated in each of these elements differs with load condition. Additional energy needed to run the control electronics and maintain charge on the implanted battery must also be provided through the inductive link. This energy expenditure is largely independent of load conditions. Data were measured for a 100 mL TAH operating on a mechanical CV system analog at 10 L/min flow rate into 14.6 kPA (110 mm Hg) pressure head (17). Of a 19.1 W energy expenditure, transcutaneous energy transmission (TETS) accounted for 25%, control electronics for 3%, the motor and its electromechanical linkage and blood pump for 67%, and kinetic energy imparted to the blood for 15%. Of particular importance to the patient is energy utilization by the implant (as distinct from the TETS), since this determines the run time provided by an implanted battery of a given technology and size, and therefore has a great impact on safety and lifestyle. Energy lost in the various components of the energy chain can be estimated given a model of each component’s losses. In accord with standard models and manufacturers’ data, we characterize the major components as follows: The electric motor and power semiconductors are described by a resistance, torque constant, constant frictional loss, and speed-dependent frictional loss, the frictional losses occurring due to nonidealities in the magnetic materials. Ball bearings are characterized by constant and speed-dependent friction terms, in accord with the standard model. The rollerscrew mechanism is characterized by an efficiency (forcedependent friction), in accord with data supplied by its manufacturer. We must recognize also that energy is dissipated in accelerating the moving parts of the system, some of which is not recovered when they decelerate at the end of each motion.


Losses in each component were estimated by assuming perfect tracking of the desired motion trajectory, and using a constant pressure to represent the circulatory load. Figure 6 shows model results for a 225 ms blood pump ejection into a mean pressure of 13.3 kPa (100 mm Hg). Torque to overcome each mechanical loss must be provided by the dc motor, and some additional resistive losses in the motor windings occur due to each of these. In this version of the model, these resistive losses are counted as part of the loss in each component. Fortunately, the largest amount of power goes to providing pressure and flow in the pump, although the sum of the losses equals the output power. No single motor component stands out as responsible for the majority of energy loss. TRANSCUTANEOUS ENERGY TRANSMISSION AND BATTERIES Currently available ventricular assist devices have successfully provided circulatory support in patients awaiting cardiac transplantation, for periods exceeding one year (18). However, these devices have all required infection-prone percutaneous access sites for the cables or tubes which power and control the pumps. Transcutaneous energy transmission systems have been under development for many years. Eventual clinical use of completely implanted systems depends not only on the development of these energy transmission systems, but more importantly on attaining the high degree of reliability in all of the implanted components necessary for long-term clinical use. In a typical application, inductive coupling between implanted and external coils transfers electrical energy across the intact skin, to power the electromechanical pumping device and control electronics, and to provide recharging power for the implanted batteries. Implanted rechargeable batteries provide backup power in the event that the external energy transmission coil is removed, and to allow the patient a brief period of tether-free operation for bathing. A wireless telemetry link may be provided by a separate coil or pair of coils located within the electronics enclosure, infrared sensors placed subcutaneously, or by modulation of the energy transmission system carrier.

661

System Requirements The design requirements for the energy transmission system, like any power supply design, include input voltage range, output voltage and regulation, output current range, and bandwidth. The input voltage is constrained by the need for a portable, battery power source, most commonly 12 V to 15 V. The output voltage is typically chosen to be higher than the implanted battery voltage and circuit supply voltages, to permit the use of simple recharging and voltage regulation circuits. Typically, an output voltage of 10 V to 15 V is required. Increasing the number of cells to achieve higher voltages results in reduced battery reliability and reduced volumetric efficiency. The design of the control circuit and battery charger usually dictate the voltage regulation requirement (Fig. 7). The output current and bandwidth vary widely, depending on the characteristics of the energy converter and the amount of energy storage provided by capacitors. While most devices require approximately 10 W mean power delivered to the implant, the dynamic power may range from 1 to 70 W during the cardiac cycle. Coil Design The inductive coupling of two coils can be described by the mutual inductance M. Mutual inductance can be calculated for two single turn circular axisymmetric coils using Neumann’s formula (19). Numerical and closed form methods for the case of angular and lateral displacements are also available (19,20). Multiturn coils are typically considered as a single coil of an estimated mean radius. Empirical coil design formulae (21) or separate Neumann solutions are also used. A useful definition of mutual inductance is √ M = k L1 L2

Power (W)

where k is the coupling coefficient, L1 is the primary coil self inductance, and L2 is the secondary coil self inductance. The coupling coefficient k is a measure of the degree of magnetic flux linkage between the coils, and ranges from zero to one. A high value of M is desirable for attaining high efficiency in power transmission applications, as it represents the best utilization of coil current. The coupling coefficient is increased by reducing the distance between the primary and secondary coil windings, or by directing the flux linkage path through 6.000 the use of magnetically permeable materials, such as ferrite. Coupling variation is expected to be common in patients 5.000 due to lateral, axial, and angular displacements of the priAoP W mary coil. Consequently, efficiency is reduced and the electri4.000 Motor W cal transfer functions of the link are affected. The significance Bearing W of transfer function variation depends upon the tolerance of a 3.000 Rollerscrew W given system to changes in output (ie, at the implant) voltage Acceleration W and current. In most designs, greater coupling coefficients are 2.000 obtained at the expense of displacement tolerance. Coil shapes which place the primary windings close to the second1.000 ary windings are inherently sensitive to misalignment. 0.000 Important factors in coil design are the ease of main0.0000 0.0500 0.1000 0.1500 0.2000 0.2500 taining alignment, patient comfort and aesthetics, minimizaTime (sec) tion of skin irritation caused by compression or abrasion, and rejection of heat from the coil windings. The secondary coil is Figure 6. Example of model calculations of losses in the energy converter power chain. Results are for a 225 ms systolic ejection into usually intended to be implanted subcutaneously, at a depth a constant 100 mm Hg load. Losses attributed to each component of 5 mm to 15 mm. An additional problem is the presence of metal (automobile body, appliance, etc.) near the coils, include I2R losses in the motor due to the component.

662


External Battery or power supply

Figure 7. Simplified diagram of typical electrical power flow to implanted pump.

which may introduce eddy current and hysteresis losses, and a shift in the coil inductances and mutual inductance (22). In this case, both coil design and drive circuit topology are important. Equidiameter planar coils achieve high coupling coefficients but are difficult to align without requiring skin pressure. A conical secondary coil and mating primary coil result in improved alignment. For circular coils of differing diameter, a degree of lateral displacement tolerance is achieved when the smaller coil remains within the circumference of the larger coil. If the inner coil is the smaller of the two, and is shaped as a raised mound, then a natural alignment is provided for the loop-shaped primary coil (23,24). Single turn primary and secondary coils encircling the abdomen have also been developed (25). Coil Network The presence of body tissue between and near the coils is of concern, both for operation of the electrical system, and for concern over the possible effects of the electromagnetic fields on the tissue. For frequencies below approximately 500 kHz, power absorption in body tissue due to currents induced by the magnetic field has been shown to be relatively minor, as has dissipation due to dielectric losses (26). Electrical modeling of the coil network and functional testing rarely require inclusion of tissue properties. Both primary and secondary coils are usually connected to capacitors in either series or in parallel to form resonant circuits in order to reduce the voltage or current levels at the input and output. Commonly, the primary and secondary coil networks are tuned to the same frequency. Operation at resonance causes cancellation of the reactances, which presents a resistive load to the driving circuit. A bipolar switching voltage drive, such as a Class D driver, is used to provide the driving voltage. Operating frequencies range from 100 kHz to 500 kHz. The transcutaneous coil pair and power switching circuits are sources of electromagnetic interference (EMI). Resonant topologies and standard techniques of EMI suppression are necessary to meet regulatory requirements. The primary and secondary networks may also be tuned to different frequencies. This may be used to compensate coupling-dependent gain variation or to increase the link bandwidth for telemetry (27). In high-power systems, stagger tuning has been applied as a means to maintain both primary and secondary resonance over changing load and coupling conditions (28). The definition of gain for a given system depends on the coil network. For example, the doubly series tuned network can be described as a voltage input, current output network

Skin

Power oscillator and controller

Internal Rectifier, filter, and regulator

Primary coil

Secondary coil

Pump controller

Blood pump

Backup battery

whose gain is the transfer admittance (Iout /Vin). The goals of both a maximum transfer admittance and a minimum coupling-dependent gain variation can be achieved by designing the network to operate at critical coupling, which is defined as the condition at which the primary circuit impedance matches the secondary impedance. However, critical coupling does not guarantee maximum efficiency when including input and output voltage constraints. CARDIOMYOPLASTY FUNDAMENTALS AND MUSCLE-POWERED BLOOD PUMP Dynamic cardiomyoplasty (CMP) is a surgical therapy for heart failure which involves the direct application of electrostimulated skeletal muscle for circulatory support. In this procedure, the latissimus dorsi (LD) muscle is isolated, wrapped around the heart, and electrically stimulated to contract during cardiac systole. The potential advantage of this approach is that long-term cardiac assist may be achieved with minimal hardware and low maintenance relative to conventional assist techniques (i.e., mechanical blood pumps). However, questions concerning the mechanism of CMP assist and the long-term efficacy of this approach remain to be answered. The use of electrically stimulated skeletal muscle as an endogenous power source offers an attractive alternative to the chronic drive systems currently in use. Muscle-powered devices have the potential to greatly simplify cardiac implants by eliminating electromechanical components and avoiding the need to transmit energy across the skin. This approach is especially appealing when one considers the substantial quality-of-life benefits to be derived from a self-contained system free from external components and daily maintenance. Moreover, the relative simplicity of such systems could drastically reduce the cost of long-term cardiac support, increasing its viability from a societal perspective. The feasibility of biomechanical circulatory support ultimately hinges on the ability of skeletal muscle to generate useful hemodynamic work on a continual basis. The persistent problem of muscle fatigue seemed to preclude such bioactuated systems until 1976 when Salmons and Sreter demonstrated that skeletal muscle could be electrically ‘‘conditioned’’ to resist fatigue (29). Since that time, a number of investigators have quantified the chronic power output of trained skeletal muscle, both in theory and via experimentation (30–33). Predictions of steady-state work capacity range from 2.0 to 15.0 mW/g of muscle tissue. Adopting the lowest figure, one can calculate that a trained muscle weighing 550 g could supply the 1.1 W required to move 5 L of blood each minute across a pressure gradient of 13.3 kPa. This


muscle mass requirement is compatible with the use of human latissimus dorsi (LD) muscle which averages 600 g in the male. However, actual power requirements will depend on the degree of circulatory support needed and the efficiency of muscle power conversion and transmission. The key to utilizing muscle power for circulatory support lies with the development of a practical scheme by which contractile energy may be collected and efficiently delivered to the bloodstream. The most popular techniques employed to date include wrapping the heart for direct mechanical assistance (cardiomyoplasty), wrapping the aorta for counterpulsation (aortomyoplasty), shaping the muscle into a neoventricle to compress a hydraulic pouch, and positioning a compressive device beneath the muscle midline. However, low power production has proven to be a major disadvantage common to all these assist schemes (34). The principal cause of this poor performance is the inherent mechanical inefficiency that results when muscle is wrapped to compress the heart or some fluid-filled conduit. Skeletal muscles contain myofibers arranged linearly to produce shortening in one direction. Therefore, muscles arranged in this manner tend to pull and twist the wrapped vessel while providing little force for centralized compression. Likewise, devices placed beneath the muscle midline access only a small fraction of the available energy because their movement is nearly perpendicular to the primary force vector of the muscle. Another important factor which limits these heterotopic methods is the functional loss brought about via muscle mobilization. Wrap-around techniques require that the muscle be isolated from all surrounding structures, sacrificing collateral blood supply and removing the muscle from its optimal orientation and stretch. The surgical isolation of skeletal muscle has been shown to produce an immediate 37% decrease in contractile power due to trauma and physical separation from surrounding synergistic musculature (35). In the chronic setting, reduced blood flow caused by the separation of collateral blood vessels further compromises function and often leads to ischemia and muscular atrophy (36). Efforts to optimize the function of wrapped muscle include the use of dynamic training techniques wherein muscles are allowed to shorten (i.e., perform work) during the conditioning process. There is evidence to suggest that skeletal muscles trained in this manner exhibit improved contractile speed and power (37). Work to use such dynamically trained muscles for cardiac support is ongoing and includes the recent development of an axisymmetric blood pump actuated via compression of a fluid-filled elastomeric bladder (38). In this scheme, LD muscle is wrapped around the bladder and stimulated to drive fluid into the pump’s housing. There, the fluid compresses a blood-filled, valved conduit and pushes blood from the left ventricle into the aorta. Given the tensile nature of skeletal muscle contraction, the most effective way to harness muscular work is to employ a linear geometry with the muscle tendon detached from its original insertion and reconnected to a hydraulic energy transmission device. Such an arrangement offers a three-fold advantage in that natural contractile mechanics are preserved, blood flow to the muscle is maintained, and power transfer efficiency is optimized. Early attempts to harvest in situ skeletal muscle for cardiac assist have employed an assortment of mechanisms. The

663

first attempt to power a pump with linearly contracting muscle was made in 1964 by Kusserow and Clapp who used a canine quadriceps femoris to actuate a levered extracorporeal pump (39). Since that time, several reports have been published describing various means to collect energy from linear muscle contractions, but relatively little emphasis has been placed on the detailed engineering needed to reduce these concepts to practice (40,41,42). As a result, these efforts have failed to produce a practical means by which contractile energy may be collected and transmitted in vivo to perform work within the body. Current work focuses on eliminating the shortcomings of previous devices in order to produce a practical muscle energy converter (MEC) for long-term use. Recently, significant progress has been made by Trumble and Magovern with the manufacture of a prototype device designed to transform in situ muscle contractions into hydraulic power (43). This MEC, shown in Fig. 8, resembles a piston pump and is designed for implantation beneath the humeral insertion of the LD muscle. Viscous and inertial losses are minimized by transmitting hydraulic energy under conditions of high pressure and low flow. Short stroke lengths (1 cm) are employed to optimize device durability and minimize trauma to surrounding tissues. Preliminary in vitro testing has demonstrated ⬎98% efficiency in converting input power to hydraulic energy and preload work. These results show that a significant amount of contractile energy can be efficiently transformed to hydraulic power via this mechanism. Should MEC implant trials prove successful, this device could be coupled to a hydraulic blood pump to form a permanent muscle-actuated ventricular assist system (MAVAS) free of all external hardware. Such technology would provide a rel-

Outer bellows

Linear bearing

Inner bellows

Outlet port

Figure 8. Left: Cross-sectional drawing of the MEC and its internal components (shown with the piston fully compressed). Right: Artist’s conception of the MEC implanted beneath the latissimus dorsi muscle. The device is anchored to the ribcage via two titanium plates which fit into a groove machined into the cylindrical housing. Muscle contractions are controlled via an implanted stimulator (shown beneath the sternum) which delivers bursts of electrical impulses to the thoracodorsal nerve.

664


atively inexpensive alternative to heart transplantation and enable patients to retain a high quality of life. Alternate applications for such bio-actuated power supplies might also include driving respiratory support devices, providing sphincter control, and actuation of prosthetic limbs. OTHER ARTIFICIAL ORGANS Implanted artificial organs have a wide range of energy and control requirements. Some need no external energy, for example prosthetic knees, hips, or cardiovascular stents. Others have modest energy needs which can be met for years by implanted batteries. These include the pacemaker and the cardioverter/defibrillator. The energy requirements for implanted cardiac pacemakers led to early designs for transcutaneously rechargeable systems, and spurred the development of lithium–iodine primary battery technology (44), the battery system still in general use today. Mechanical blood pumps, because of the substantial work they perform, cannot be powered for long with existing battery systems. Musclepowered prostheses have the potential of providing high-energy output with only stimulus, command, and control functions, and these can be accommodated with implanted batteries. Implanted drug or insulin delivery systems are examples of fluid-pump artificial organs which do not require large amounts of energy to function. Functional Electrical Stimulation of Muscle There are development efforts underway on systems intended to stimulate skeletal musculature to achieve postural maintenance (45) and even ambulation (46) for patients without voluntary muscle control, such as those with major damage to the spinal cord. In effect, the controller for these systems replaces an element of coordination normally performed by the spinal cord, brainstem, and cerebellum. One system undergoing clinical trials (47) provides hand grasp using electrodes implanted along and within the muscles of the forearm. Control is provided by shoulder muscle activation. These systems are designed to utilize the power available from the patients’s own muscles. One limitation on these system designs is the restricted dynamic range in muscle force obtained from electrically stimulated muscle. The fine gradation of muscle force obtained by neural control is lost when stimulating muscle electrically, and rapid fatigue of muscles is the result. A second problem is the development of suitable artificial transducers for force and for position to control muscle-powered systems. The hope is that such designs for muscle stimulus may eventually be integrated with position and force information obtained from the physiological sensors to produce effective closed-loop controlled motion (48). Artificial Pancreas Diabetes is a condition where the pancreas has lost its ability to regulate blood glucose levels through the release of insulin into the bloodstream. The resulting elevated blood-sugar levels have a number of serious consequences, including blindness, deficient wound-healing, and cardiovascular disease leading to early death. The lack of available organs and the need for antirejection drugs in pancreas recipients has limited

the use of pancreas transplantation. As is the case with blood pumps, there are total artificial and hybrid approaches to the design of the artificial pancreas, and control system implementation is an important area of ongoing research. Important hybrid systems undergoing development include the use of immunoisolated islets of Langerhans, the functional multicellular unit of the pancreas. These may be obtained either from human or from animal sources. These cells have the ability to naturally self-regulate insulin output in response to changes in blood sugar levels. Development efforts are directed at producing membranes which allow inward diffusion of glucose and oxygen and outward diffusion of insulin, while blocking elements of the immune system which would destroy the islets (49). The total artificial approach to the artificial pancreas includes the clinically important externally worn infusion pump (eg, MiniMed 507, MiniMed, Fridley, Minnesota), in use by tens of thousands of patients. This device pumps insulin through a Teflon tube into the abdominal cavity at rates of up to about 1 애L/min. Worn externally it is readily refilled with insulin and batteries can be replaced as needed. Attempts to apply closed-loop control based on blood-glucose sensing have generally been unsuccessful. Sensor reliability remains a significant problem. Externally worn pumps run in an open-loop control configuration, with programmable basal rate, and elevated doses provided at mealtimes. A fully implantable intra-abdominal insulin pump is a long-term goal of continuing research efforts (50), and at least one design has been introduced for clinical use (Minimed MMT-2001, MiniMed, Fridley, Minnesota). This design eliminates the need for a percutaneous Teflon tube leading to the abdominal cavity. Periodic recharging of the insulin reservoir via hypodermic would be required. Lithium battery replacement would be needed at intervals of 3 years or more. In the case of the insulin pump, the flow rates are very low, and the pressure load within the abdomen is modest, so that the hydraulic workload is negligible. However, the electronics and pump energy requirements limit battery life. One advantage of the implanted insulin pump is that it may eventually provide a better location for stable glucose sensor operation and so enable fully closed-loop control for the artificial pancreas. Support was received during the preparation of this section from NIH contracts N01-HV-38130 and N01-HV-58156 and by a grant from the Whitaker Foundation. BIBLIOGRAPHY Design and Control of Blood Pumps 1. R. T. V. Kung et al., An atrial hydraulic shunt in a total artificial heart—A balance mechanism for the bronchial shunt, Amer. Soc. Artif. Intern. Organs J., 39: M213–M217, 1993. 2. A. J. Snyder et al., In vivo testing of a completely implanted total artificial heart system, Amer. Soc. Artif. Intern. Organs J., 39: M177–M184, 1993. 3. R. Rintoul et al., Continuing development of the Cleveland Clinic—Nimbus total artificial heart, Amer. Soc. Artif. Intern. Organs J., 39: M168–M171, 1993. 4. H. C. Kim et al., Development of a microcontroller-based automatic control system for the electrohydraulic total artificial heart, IEEE Trans., Biomed. Eng. 44: 77–89, 1997.

ARTIFICIAL HEARTS AND OTHER ORGANS 5. H. Konishi et al., Long-term animal survival with an implantable axial flow pump as a left ventricular assist device, Artif. Organs, 20 (2): 124–127, 1996. 6. R. J. Kaplon et al., Miniature axial flow pump for ventricular assistance in children and small adults, J. Thorac. Cardiovasc. Surg., 111 (1): 13–18, 1996. 7. L. A. R. Golding and W. A. Smith, Cleveland Clinic rotodynamic pump, Ann. Thorac. Surg., 61: 457–462, 1996. 8. K. Kawahito et al., Ex vivo evaluation of the NASA/DeBakey axial flow ventricular assist device, Amer. Soc. Artif. Intern. Organs J., 42 (5): M754–757, 1996. 9. A. J. Snyder, G. Rosenberg, and W. S. Pierce, Noninvasive control of cardiac output for alternately ejecting dual-pusherplate pumps, Artif. Organs, 16: 182–194, 1992. 10. A. J. Snyder, G. Rosenberg, and D. L. Landis, Indirect estimation of circulatory pressures for control of an electric motor driven total artificial heart, Adv. Bioeng., 87–88, 1985. 11. J. F. Gardner et al., Aortic pressure estimation with electro-mechanical circulatory assist devices, J. Biomech. Eng., 115: 187– 194, 1993. 12. D. B. Geselowitz, G. E. Miller, and W. M. Phillips, Dynamic mode of a c-type pneumatically driven artificial ventricle, J. Biomech. Eng., 14–19, 1977. 13. T. Kitamura, T. Kijima, and H. Akashi, Modeling technique of prosthetic heart valves, J. Biomech. Eng., 106: 83–88, 1984. 14. G. Rosenberg et al., Design of and evaluation of the Penn State mock circulatory system, Am. Soc. Artif. Organs J., 4: 41–49, 1981. 15. N. C. Kelley et al., Noninvasive control of the Penn State total artificial heart: A computer simulation, Proc. Northeast Bioeng. Conf., 22–23, March 1995. 16. L. Martin et al., Non-invasive control of aortic pressure and cardiac output in a reciprocating pusher-plate artificial heart, Proc. Northeast Bioeng. Conf., pp. 62–63, March 1996. 17. G. Rosenberg et al., Power requirements for an electric motordriven total artificial heart, Proc. IEEE 9th Annu. Conf. Eng. Med. Biol. Soc., 1987, pp. 188–189. 18. B. P. Griffith et al., Results of extended bridge to transplantation: Window into the future of permanent ventricular assist devices, Ann. Thorac. Surg., 61: 396–398, 1996. Transcutaneous Energy Transmission 19. F. C. Flack, E. D. James, and C. M. Schlapp, Mutual inductance of air-cored coils: Effect on design of radio-frequency coupled implants. Med. Biol. Eng., 9: 79–85, 1971. 20. M. Soma, D. Galbraith, and R. L. White, Radio-frequency coils in implantable devices: Misalignment analysis and design procedure, IEEE Trans. Biomed. Eng., 34: 276–282, 1987. 21. F. E. Terman, Radio Engineering, 2nd ed., New York: McGrawHill, 1937. 22. D. B. Geselowitz, Q. T. N. Hoang, and R. P. Gaumond, The effects of metals on a transcutaneous energy transmission system, IEEE Trans. Biomed. Eng., 39: 928–934, 1992. 23. C. Sherman et al., Research and development: Systems for transmitting energy through intact skin, Final Technical Report N01HV-0-2903-3, Thermo Electron Corp., Waltham, MA, July 1983. 24. W. J. Weiss et al., In vivo performance of a transcutaneous energy transmission system with the Penn State motor driven ventricular assist device, Trans. Am. Soc. Artif. Intern. Organs, 35 (3): 284–288, 1989. 25. J. S. Brugler et al., Transcutaneous power transmission and electronic control of a ventricular assist system, Proc. IEEE 8th Annu. Conf. Eng. Med. Biol. Soc., 73–76, 1986.

665

26. J. C. Schuder, J. H. Gold, and H. E. Stephenson, Jr., An inductively coupled rf system for the transmission of 1 kW of power through the skin, IEEE Trans. Biomed. Eng., BME-18: 265– 273, 1971. 27. D. C. Galbraith, M. Soma, and R. L. White, A wide-band efficient inductive transdermal power and data link with coupling insensitive gain, IEEE Trans. Biomed. Eng., 34: 265–275, 1987. 28. J. A. Miller, G. Belanger, and T. Mussivand, Development of an autotuned transcutaneous energy transfer system, Amer. Soc. Artif. Intern. Organs, J., 39 (3): M706–M710, 1993. Muscle-Powered Blood Pumps 29. S. Salmons and F. Sreter, Significance of impulse activity in the transformation of skeletal muscle type, Nature, 263: 30–34, 1976. 30. D. R. Trumble and J. A. Magovern, Ergometric studies of untrained skeletal muscle demonstrate feasibility of muscle-powered cardiac assistance, J. Appl. Physiol., 77 (4): 2036–2041, 1994. 31. F. Ugolini, Skeletal muscle for artificial heart drive: Theory and in vivo experiments. In: Biomechanical Cardiac Assist: Cardiomyoplasty and Muscle-Powered Devices, Mount Kisko, NY: Futura, 1986, pp. 193–210. 32. S. Salmons and J. C. Jarvis, The working capacity of skeletal muscle transformed for use in a cardiac assist role. In: Transformed Muscle for Cardiac Assist and Repair, Mount Kisko, NY: Futura, 1990, pp. 89–104. 33. J. C. Jarvis, Power production and working capacity of rabbit tibialis anterior muscles after chronic electrical stimulation at 10 Hz, J. Physiol. (Great Britain ) 470: 157–169, 1993. 34. S. Salmons and J. C. Jarvis, Cardiac assistance from skeletal muscle: A critical appraisal of the various approaches, Br. Heart J., 68: 333–338, 1992. 35. W. N. Stainsby and G. M. Andrew, Maximal blood flow and power output of dog muscle in situ, Med. Sci. Sports Exercise, 20: S109– S112, 1988. 36. C. A. Doorn et al., Latissimus dorsi muscle flow during synchronized contraction: Implications for cardiomyoplasty, Ann. Thor. Surg., 61: 603–609, 1996. 37. N. W. Guldner et al., Dynamic training of skeletal muscle ventricles: A method to increase muscular power for cardiac assistance, Circulation, 89: 1032–1040, 1994. 38. R. L. Whalen et al., A skeletal muscle powered ventricular assist device (BAD), Amer. Soc. Artif. Intern. Organs, J. Abstracts, 42 (2): 132, 1996. 39. B. Kusserow and J. Clapp, A small ventricle-like pump for prolonged perfusions: Construction and initial studies, including attempts to power a pump biologically with skeletal muscle, Trans. Amer. Soc. Artif. Intern. Organs, 10: 74–78, 1964. 40. D. Spitzer, An implantable power source for an artificial heart or left ventricular assist device, Trans. Amer. Soc. Artif. Intern. Organs, 31: 193–195, 1985. 41. D. J. Farrar and J. D. Hill, A new skeletal linear-pull energy convertor as a power source for prosthetic circulatory support devices, J. Heart Lung Transplant, 11: S341–S350, 1992. 42. M. Takahashi et al., Efficacy of a skeletal muscle-powered dynamic patch: Part 1. Left ventricular assistance, Ann. Thor. Surg., 59: 305–312, 1995. 43. D. R. Trumble and J. A. Magovern, A permanent prosthesis for converting in situ muscle contractions into hydraulic power for cardiac assist, J. Appl. Physiol., 82 (5): 1704–1711, 1997.

666

ARTIFICIAL INTELLIGENCE

Other Artificial Organs 44. W. Greatbach et al., The solid state lithium battery, IEEE Trans. Biomed. Eng., BME-18: 317, 1971. 45. N. de N. Donaldson and C.-H. Yu, FES standing: Control by handle reactions of leg muscle stimulation (CHRELMS), IEEE Trans. Rehabil. Eng., 4: 280–287, 1997. 46. D. A. Winter, Biomechanics and Motor Control of Human Movement, New York: Wiley Interscience, 1990. 47. K. S. Wuolle et al., Development of a quantitative hand grasp and release test for patients with tetraplegia using a hand neuroprosthesis, J. Hand Surgery, 19A: 209–218, 1994. 48. T. R. D. Scott, P. H. Peckham, and K. L. Kilgore, Tri-state myoelectric control of bilateral upper extremity neuroprostheses for tetrapleic individuals, IEEE Trans. Rehabil. Eng., 4: 251–263, 1997. 49. C. K. Colton and E. S. Avgoustiniatos, Bioengineering in development of the hybrid artificial pancreas, Trans. ASME (Biomechanics), 113: 152–170, 1991. 50. T. Buchwald, D. Rhode, and K. Kernstine, Insulin delivery by implanted pump: A chronic treatment for diabetes, Trans. Am. Soc. Artif. Intern. Organs, 35: 5–7, 1989.

ROGER P. GAUMOND JOHN F. GARDNER ALAN J. SNYDER WILLIAM J. WEISS Penn State University

CHRISTOPHER KELLEY Boston University School of Medicine

DENNIS R. TRUMBLE Allegheny University of the Health Sciences (Allegheny Campus)


}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP


Wiley Encyclopedia of Electrical and Electronics Engineering Artificial Limbs Standard Article J. E. Sanders1, S. G. Zachariah1, B. J. Hafner1 1University of Washington, Seattle, WA Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6601 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (325K)

●

●

● ●




Abstract The sections in this article are Lower Limb Prosthetics Upper Limb Prosthetics Current Research About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.

file:///N|/000000/0WILEY%20ENCYCLOPEDIA%20OF%20ELECT...NEERING/51.%20Rehabilitation%20Engineering/W6601.htm15.06.2008 19:56:07

❍

❍ ❍


692

ARTIFICIAL LIMBS

ARTIFICIAL LIMBS Artificial limbs are man-made devices intended to replace amputated or congenitally deformed feet, legs, hands, or arms. The main purpose of an artificial limb is to replace function, to mechanically replace the part of the extremity that no longer exists. Another goal is to provide a cosmetic appearance similar to a normal limb. The earliest surviving lower limb prosthesis is dated at approximately 300 B.C. Used in the Samnite Wars in Capri, Italy, the prosthesis was made of bronze and wood and was shaped to resemble the thigh, knee, and calf. It functioned to replace the missing extremity on an active lower limb amputee. However, a written report of the use of an artificial limb was documented over 100 years earlier (1). Hegistratus of Elis, a seer who was condemned to death in 424 B.C. by the J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

ARTIFICIAL LIMBS

Figure 1. (left) A transfemoral prosthesis invented by Ambroise Pare in the mid-1500s. (From A. Pare: Oeuvres Completes, Paris, 1840. From the copy in the National Library of Medicine. From G. T. Sanders, B. J. May, R. Hurd, and J. Milani, Lower Limb Amputations: A Guide to Rehabilitation, Philadelphia: F.A. David Company, 1986, with permission.) (right) An iron artificial left hand and arm from approximately 1602. (From M. Vitali, K. P. Robinson, B. G. Andrews, E. E. Harris, and R. G. Redhead, Amputations and Prostheses, 2nd ed., London: Bailliere Tindall, 1986, with permission.)

Spartans and tethered by his leg while awaiting execution, amputated his foot to escape. He traveled 30 miles to Tregea. However, in Zaccynthius, he was again captured by the Spartans who this time successfully executed him. Their records indicate that he wore a wooden foot at the time of his death. Early lower limb prostheses, though bulky and inefficient by today’s standards, bear some design features similar to modern-day artificial limbs [Fig. 1(left)]. One of the first reported above-knee prostheses was described by Ambrose Pare in 1564. It utilized fixed equinus (fixed plantarflexion) and a controlled knee lock, features still found in some modern prosthetic designs. In 1696 Verdiun produced a below-knee prosthesis with a leather socket and thigh corset with articulated side steels to hold the prosthesis on and stabilize it with the residual limb, a design similar to that used in the twentieth century. The ‘‘Bly’’ leg, patented in 1858, included a functional ankle. It allowed plantar and dorsiflexion and also lateral motion. In 1860 Marks substituted a hard rubber foot for a wooden foot, creating a more dynamically active prosthetic foot, a concept introduced into a number of commercial products in the latter half of the twentieth century. During the Civil War, Hanger, an amputee in the Confederate army, produced the first articulated prosthetic feet by placing rubber bumpers within solid feet designs. In the twentieth century, war pushed prosthetic advances further. As a result of a need for fitting World War II veterans, the Veterans’ Administration supported development of

693

two new socket designs: the patellar-tendon-bearing socket for below-knee amputees, which was designed to apply much of the weight-bearing load on the durable patellar tendon immediately below the knee cap; and the quadrilateral socket for above-knee amputees, which transferred the majority of the weight-bearing load directly to the ischium and ensured that the position of the prosthetic brim was maintained with respect to the ischium. Advances continued in the 1960s with more frequent use of endoskeletal prostheses, which have a central post through which the force is transferred, as opposed to exoskeletal units, which are hollow with an external frame. Endoskeletal prostheses have the advantages of modularity and weight reduction. The history of upper limb prosthesis [Fig. 1(right)] also dates back more than 2000 years. The first report of an artificial hand is from the Second Punic War (218 to 202 B.C.), where Marius Sergius, a soldier, lost his right hand during battle and was subsequently fitted with an iron hand (2). The Alt–Ruppin hand discovered in 1800 dates to approximately 1400 and was made of iron with a rigid thumb fixed in opposition. Flexible fingers operating in pairs could be flexed passively and locked into position with a ratchet. Similar to this design was one of the best known artificial hands, that of a German knight, Gotz von Berlichingen who lost his hand at the siege of Landshut in 1509, as described in a poem by Goethe. The fingers could be flexed passively and locked into position with a ratchet. Prehension, harnessing of the shoulder girdle muscles to allow shoulder motion to control function of the terminal device, was an enhancement introduced to upper limb prostheses in the nineteenth century in Berlin, Germany. Initial systems were reported devised by a Berlin dentist, Peter Ballif (2). However, prehension in those designs was used only for below-elbow amputation. Subsequently in 1844 a Dutchman, Van Peetersen, used the same mechanism to achieve elbow flexion. In 1855 a design that allowed pressure on a lever against the chest to induce elbow flexion was described. The post-Vietnam War era saw the introduction of myoelectric control, a method by which neural signals from the residual limb are used to control externally powered devices, further enhancing the ease of control of upper extremity prostheses. The scope of amputation provides insight into the populations for whom prostheses must be designed. As of 1984 there were approximately 400,000 amputees in the United States with approximately 60,000 new amputations performed each year (3). Principal reasons for amputation include severe injury, disease (e.g., cancer, diabetes), and congenital defects. Traumatic injury and vascular-related diseases are the principal causes. Approximately 58% of new amputations are on patients between the ages of 21 and 65. Thus there is a significant patient population of young people with amputations, a group likely to conduct strenuous activities when using their prosthetic limbs. For persons over the age of 50, vascular causes are the etiology in 89% of the cases. These individuals typically, but not always, seek a prosthesis that simply allows function or provides a cosmesis. Artificial limbs are classified on the basis of the number of intact joints proximal to the level of amputation. For the lower-limb, amputation levels include partial foot, syme, transtibial (below-knee), knee disarticulation (through-knee), transfemoral (above-knee), and hip-disarticulation (Fig. 2). Those for upper limb include partial hand/wrist disar-

694

ARTIFICIAL LIMBS

Shoulder disarticulation

Hip disarticulation Short above-knee

Short transhumeral

Standard transhumeral Elbow disarticulation

Medium transfemoral Long transfemoral Supracondylar Short transtibial

Very short transradial Short transradial

Standard transtibial

Long transradial Long transtibial Wrist disarticulation Partial hand

Syme Partial foot

Figure 2. Levels of lower limb and upper limb amputation.

ticulation, transradial (below-elbow), elbow disarticulation (through-elbow), transhumeral (above-elbow), and shoulder disarticulation. In general, a surgeon performing an amputation tries to save as many joints as possible so as to maximize function while still overcoming the etiology that required the amputation. Most artificial limbs attach to the residual limb via a socket and supplementary apparatus. The socket is usually custom-made for the individual whereas the remaining components, for example the foot or hand, are off-the-shelf items. The inside socket shape is typically not an exact copy of the shape of the residual limb but is instead a modified shape pushed into the residual limb in load-tolerant and supportive regions whereas it is off-loaded in sensitive areas. Supplementary apparatus to hold the prosthesis onto the limb may include straps, sleeves, or cables. Most artificial limbs are passive devices, that is, they are made of deformable materials and are controlled by the musculature of the residual limb. Energy-storage-and-return, lower limb componentry (feet, ankles), introduced in the 1980s, helped to enhance the efficiency of passive artificial legs. Energy storage components are made of deformable materials that store energy internally when stressed and return that energy as they are unloaded, springing back to their original shape. Active artificial limbs exist, though they are used mainly in upper limb applications, in part, because upper limb prostheses are not subjected to as high levels of load bearing, have lower power requirements, and require finer control than lower limb prostheses. LOWER LIMB PROSTHETICS Amputations of the lower limb account for approximately 80% to 85% of all amputations performed annually in the United States. The principal cause is dysvascular disease. Amputation is often the optimal solution for a painful dysvascular limb. Approximately 80% of amputations for dysvascular disease occur in patients with diabetes (4). Other reasons for

causes of lower limb amputation include traumatic injury, tumors, and congenital defects. The primary purpose of a lower limb prosthesis is to provide functional ambulation. Because amputational surgery and prosthetic fitting are geared toward this goal, electronic componentry is used less often than in upper limb applications. Further, research and development efforts focus on creating lightweight, strong, energy-returning components that enhance ambulatory efficiency. Amputational Surgery The goals of lower limb amputational surgery are to remove a section of a limb so as to eliminate a pathological state and to create a residual limb that permits functional ambulation when fitted with an appropriately prescribed prosthesis. In a traumatic injury, often the surgeon must make do with the residual limb tissues that remain, trying to save bone length if the residuum is short but ensuring sufficient viable soft tissue for covering. A very short bone length provides insufficient residual limb surface area for load bearing and stability. A joint with much adherent scar tissue is also difficult to manage because of frequent soft tissue trauma. Thus a joint might be sacrificed in these cases. In a nontraumatic situation (e.g., amputation due to peripheral vascular disease), the surgeon can be more consistent. With the posterior flap surgical technique, skin and soft tissue covering the gastrocnemius and soleus muscles is pulled around the distal end of the tibia and fibula and sutured to the anterior tibial surface. This technique ensures that the well-vascularized posterior tissues cover the distal end of the residual limb (5). Whether amputation is for a traumatic or nontraumatic reason, soft tissues are handled in the gentlest way possible. Vessels are double-ligated and cut at the level of amputation. To avoid painful neuromas, nerves are pulled gently, resected at a sharp angle with a sharp blade proximal to the level of amputation, then released and allowed to retract back into the wound. The anterior-distal tibia is rounded with a file so as to reduce soft tissue trauma over the distal end of the bone. The fibula is typically cut 1 cm proximal to the tibia so as to avoid soft tissue trauma at its distal end. The wound is usually closed in layers, first the fascia, and then the skin. A suction drain is often used postoperatively for one to two days to ensure proper drainage of wound fluid. The wound is covered with wool and plaster of Paris dressing before the early prosthetic fitting begins. Design of a residual limb to permit functional ambulation represents a challenging biomechanical effort. The aim is to use the remaining structures (e.g., muscles, tendons, bone) to maximal advantage, particularly to overcome anticipated fitting problems of the residual limb in the socket. For example, suspension of the residual limb in the prosthetic socket during the swing phase is enhanced by suturing muscles in slight tension so that the activated muscles cause an enlarged diameter residual limb, helping to hold the prosthesis on the residual limb during the swing phase. A bone bridge, a bony graft inserted between the tibia and fibula, helps to create a more stable residual-limb bony structure. Types of Prostheses The nature of the amputation, in part, determines the type of prosthesis. For transtibial amputations, the more common

ARTIFICIAL LIMBS

types of prostheses include the total contact, patellar-tendonbearing (PTB) design in which much of the load is tolerated on the durable patellar tendon distal to the knee cap. PTB sockets are very much the standard for transtibial amputation. The main variations are in the methods of suspension (e.g., y-strap, supracondylar, cuff, supracondylar medial wedge, and neoprene sleeve). A suprapatellar socket design is a variation where the proximal brim goes over the patella and hooks proximally, thus aiding suspension. Transfemoral socket design is an area of active development. Quadrilateral sockets developed in the early 1950s transfer the majority of load bearing to the ischium. The ischial-containment socket gradually replaced the quadrilateral socket, transferring load to the ischium but distributing more of it to the gluteus, femur, and soft tissues of the residuum. A subsequent variation was to introduce an inner flexible socket that allows some shape adaptability during muscle contraction, improved heat transfer, sensation, and suspension due to greater contact friction. An outer carbon-fiber frame allows considerable weight reduction, and also large window openings in the socket wall allow the flexible socket room to move when adapting to shape. Channel-shaped openings contain and orient active muscle groups. Prosthetic Design and Fitting Fabrication of an artificial limb typically involves three major stages: custom fabrication of the prosthetic socket; selection of off-the-shelf components; and assembly and alignment of the complete prosthesis. Each stage calls for considerable skill and experience from the prosthetist. The prosthetist must take into account the nature of the individual residual limb, the lifestyle of the amputee, and the amputee’s physical and financial ability, so as to select the basic socket design and prosthetic components that will return the greatest degree of function to the amputee. Socket Fabrication. Custom fabrication of the prosthetic socket is divided into four steps: recording the shape of the residual limb, designing the shape of the socket, fabricating the socket liner, and fabricating the socket shell. Together they form the most time-consuming and labor-intensive stage of creating an artificial limb, and thus are the most expensive. To design a prosthetic socket, first it is necessary to measure the shape of the residual limb so as to have a starting point for design. The most commonly used method for measuring shape is to make a negative mold from a wrap-cast of the residual limb. The prosthetist wraps a plaster of Paris bandage around the residual limb and then applies pressure while the bandage hardens in selective locations where the tissues are more load-tolerant. The pressure is used to distort the cast into a shape desired during actual use of the prosthesis. Use of noncontact scanning methods based on laser or patterned light are becoming more popular and are advantageous because they provide a digital record of the residual limb shape that can subsequently be used for computer-aided design and manufacturing as well as archiving. Designing the final shape of the prosthetic socket from the plaster cast or a digital image of the residual limb is called socket rectification. The goal of socket rectification is to shield sensitive soft tissues from painful or injurious stresses, while

695

Figure 3. Software used for prosthetic socket shape modification. A transfemoral socket showing rectification regions is shown using ShapeMaker software (Seattle Limb Systems, Inc., Poulsbo, WA).

transferring external load to the soft tissues that are more load-tolerant. The exact magnitude and location of the rectifications depend on many factors, including the type of socket design, the presence of scar tissues, the maturity of the residual limb, and the bulk and deformability of the soft tissues. Rectification may also provide for suspension of the socket from the residual limb and for almost complete pressure relief on the distal end. If the socket is being rectified manually, as is commonly the case, the wrap-cast is filled with plaster of Paris paste or similar casting material to produce a positive replica of the residual limb. The prosthetist manually sculpts this replica, shaving away material where directed loading of the soft tissues is required, and adding extra material where the soft tissues need to be shielded from excessive stress. It is common that transparent check sockets are made to verify appropriateness of the socket rectifications directly on the amputee’s residual limb. If a digital image is to be modified, rectification is performed with custom software for this purpose (6) (Fig. 3). When the prosthetist has finished rectifying the digital limb, a numerically controlled lathe carves the rectified positive from a blank. Computer-assisted socket rectification is fast, allows rapid fabrication of duplicates, and enables a degree of expertise to be passed to the user of the software. Though computer methods of design and manufacturing have had impact, the technologies are still in their nascent stages. Display and manipulation of three-dimensional shapes on two-dimensional devices have yet to provide the prosthetist with information about the underlying skeletal structure and the deformability (material properties) of the soft tissues as does manual handling of the tissues. This is an important drawback of current computer-aided design methods that needs to be overcome. Most socket designs for amputations below the knee incorporate a cushioning liner between the residual limb and the socket shell. The socket liner attenuates stress concentrations that occur where the bony skeleton is close to the skin surface and accommodates some of the variation in the shape of the residual limb that occurs over time. Because the liner functions as a glove over the residual limb, the prosthetist typically fabricates it directly onto the rectified positive replica of

696

ARTIFICIAL LIMBS

Figure 4. Lower limb prosthetic componentry. Two pylons (an aluminum shank and a hydraulic shank), two feet (a SACH foot and a Seattle foot that has been sectioned to show the plastic leaf spring), a heel wedge, and an assembled transradial amputee prosthesis are shown.

the limb, selecting from a number of different elastomeric foams and gels available within the industry. Silicone gel was initially selected as a liner material in transtibial socket design because of its ability to distribute shear stresses. Currently, a number of nonsilicone formulations with similar properties are available. A parallel unrelated development was suction suspension (7), made possible by the newer flexible socket materials that allow a greater degree of contact and thus better interfacial stress distribution. Elastomeric liner sleeves with a distal locking pin in the socket are used. Cushioning liners are not as common in cases where the amputation is above the knee. The final stage in the custom fabrication of the socket involves forming the load-transmitting structural shell directly over the positive replica with the liner in place. If other components, such as suction valves and inflatable bladders, are incorporated within the socket, then dummies of their shape are also affixed to the positive form. Advances in plastics technologies in recent years have allowed forming lightweight sockets from thermoplastics, such as polyethylene and polypropylene, or from thermoset polyester and epoxy resins reinforced with glass and carbon-fiber fabrics. Off-the-Shelf Components. The prosthetist selects a number of off-the-shelf components to attach to the socket to form the complete functional artificial limb (Fig. 4). The most important of these are the footpiece at the terminal end and an artificial joint or joints if required. In response to consumer demand from physically active amputees, particularly since World War II, designs have evolved considerably. A wide selection of connecting elements, adapters, and alignment devices to connect the socket and joints to the footpiece are available, as well as a variety of methods of suspension to keep the artificial limb from falling off the residual limb during walking and other activities.

The joints of the normal leg that may need to be replaced in an artificial leg are the ankle, knee joint, and hip joint. Of these, the knee joint has received most design attention. Among its passive functions are locking in extension during standing and flexing as required when seated. Active functions include absorbing shock without buckling, lengthening the limb during stance so as to accelerate the body forward, shortening the limb while it is swinging through the air and then extending it again before the leading foot contacts the ground. Effective designs achieve these functions while satisfying weight, size, and external power restrictions inherent in all artificial limb designs. Knee joint designs can be conceptually classified based on the way they control rate-dependent aspects of joint flexion. The rotational stiffness (impedance) of the joint must be very high as the leading foot contacts the ground in normal walking or during stumbling. As the knee is flexed further as part of the normal stride, the rotational stiffness needs to decrease so that the rate of rotation matches that of the opposite knee for the current walking speed. Similarly, as the foot is lifted off the ground and swung forward, the knee must first flex and then extend at the same rate as the opposite knee. Thus when the knee is bearing load (the stance phase), it must have two different stiffnesses in flexion, and when it is not bearing load (the swing phase), it must have one rate of flexion and another rate of extension. Each of these rates, in turn, should depend on the walking or running speed selected by the amputee. The simplest knee mechanisms, often called constant friction, use friction within the joint to control the rate of rotation in flexion and extension. In more sophisticated designs, called stance controlled, (8), the coefficient of friction is controlled on the basis of the amount of load borne by the leg. Designs based on friction, however, cannot respond to variations in walking speed. In the more complex, popular, and expensive knee mechanisms, pneumatic and hydraulic piston-cylinder combinations allow the designer to tightly control each portion of the walking cycle, producing in many transfemoral amputees a gait that appears entirely natural to all but the highly trained eye. Though these fluid-controlled knees successfully adapt to variations in walking speed, they are essentially tuned to work optimally about a predetermined preferred speed. To allow the active amputee an even greater range of walking and running speeds, designers have recently begun incorporating active controls on the valves that the prosthetist or amputee can set manually. These controls are mediated by microprocessors, and the manual control involves typing instructions via a detachable keypad. Devices are commercially available which sense the walking cadence and adjust the parameters of the hydraulic or pneumatic cylinders accordingly. The next step in development will be to control the valves in real time by microprocessors that use measured rate and direction of rotation as their active inputs. The role of the normal human foot–ankle complex in walking is similar in many respects to that outlined for the knee joint. As the foot contacts the ground, the ankle joint acts as a shock absorber until the foot is planted flat on the ground. As load is transferred to the foot, it must adapt to any unevenness of the terrain so as to provide a stable base on which to bear load and from which to accelerate the body forward. Finally, as the amputee prepares to take the next step, the ankle joint actively extends, contributing to the accelerating

ARTIFICIAL LIMBS

force. In addition to these dynamic requirements, there is often a cosmetic requirement that the footpiece easily fit within a shoe or other routine footwear. Despite this conceptual similarity of function between the foot and the knee, the approach to artificial ankle and footpiece design is markedly different from that of the knee joint. Whereas the knee joint relies on very intricate and complex linkage systems, in general, the footpiece is a single-unit composite of different metallic or polymeric materials without any moving parts. Because the footpiece is at the end of a swinging pendulum, its weight takes on enormous significance. There is also less space available at the end of the leg for a linkage-based, ankle-joint design. Finally, the location of the foot–ankle contacting just centimeters above the ground makes mechanisms prone to damage and wear from water, dust, and other contaminants. For about three decades between the 1950s and the 1980s, the footpiece of choice was a relatively simple design called the solid ankle cushioned heel (SACH) foot. The basic design incorporates a curved wooden keel, modeled after a shoe last, and a foam heel bumper surrounded by an elastomeric sheath. The cushioned heel absorbs energy as the heel contacts the ground, the elastomeric sheath allows a small amount of adaptability to the terrain, and the curved keel allows the amputee to roll forward on the foot in a manner that simulates ankle flexion. Notably there is no active pushoff as the foot leaves the ground at the end of the stance phase. Sophistication of this basic design includes a choice of stiffness for the heel, a choice of heel heights to match different shoe designs, and an injection-molded sheath that includes toes and other cosmetic features. A conceptual variant of the SACH foot that was popular in Europe is the singleaxis foot. It incorporates a uniaxial joint at the ankle level and has a slightly different location for the heel bumper. Some multiaxial feet were also developed, but weight and durability issues limited their widespread acceptance. With the advent of injection-molded plastics, the wooden keel in the SACH foot began to be replaced with plastic keels. A favorable feature of this material is that the torsional and bending stiffness of the keel can be controlled, allowing for greater shock absorption and adaptability to uneven terrain. A number of different keels designs evolved in this class of energy-absorbing feet, though the most notable of these was the stationary ankle flexible endoskeleton (SAFE) foot. With the departure from the solid keel of the SACH design, it was only a matter of time before designers realized that the energy stored in the distortion of the keel could be returned to assist in the push-off phase of stance. The Seattle foot, developed in the mid-1980s (9) led the trend towards these energystoring, energy-returning, or dynamic elastic response feet, as they have been variously called. There remains a great deal of debate whether these feet actually assist in the push-off phase, but that does not detract from the design philosophy which is dynamic as opposed to passive. As with the knee joint, however, a design that attempts to provide rate control of rotation performs best at a predetermined speed and can be awkward to use at other speeds. Assembly and Alignment of Components and Socket. The final stage in fabricating the prosthesis entails assembling and aligning the socket shell and the off-the-shelf components into a functional limb. The orientation of the socket with respect

697

to the footpiece is set on the workshop bench following the general recommendations of the design selected. Finer adjustments to the alignment are made through adjustment screws incorporated into the components, after observing the amputee standing or walking on the artificial limb. The goals of the alignment process are to produce a symmetric, smooth, and cosmetically acceptable walking pattern that demands a limited amount of energy from the amputee, and does not produce discomfort or pain. Comments from the amputee on the ease of walking and other aspects can be highly relevant in the alignment process, particularly for an amputee who has received several prostheses during his or her lifetime and thus has much experience. The assessment of the amputee’s walking pattern is commonly called clinical gait analysis. In its most basic form, clinical gait analysis consists of careful observation by a highly trained observer and interpretation based on an understanding of the biomechanics of amputee walking. The use of video recording and slow-motion replay can be of considerable assistance to novices as they build up their expertise. However, an understanding of prosthetic biomechanics is still required to correlate the abnormalities in the gait with specific misalignments and socket deficiencies. It is also possible to quantify and then evaluate an individual’s gait pattern for prosthetic alignment or computational evaluation, though this is generally done only for research purposes because of the complexity of data acquisition, analysis, and interpretation. Such quantification of gait involves recording the three-dimensional motion of the entire body with respect to time (kinematics), the forces transmitted through the limbs to the floor (kinetics), and the intensity and duration of muscle activity in the various groups that control the prosthetic limb (dynamic electromyography). Then these data are compared with known or expected patterns. The kinematic portion of the analysis involves recreating a segmented stick-figure model of the subject by tracking passive (reflective) or active (infrared emitting) markers on the subject with multiple video cameras. Software extracts the markers from the images, triangulates their positions, and then extracts relevant parameters of translational and rotational velocities and accelerations (10). Kinetic data can be added to the analysis if force transducers (called force plates) are positioned flush with the floor in the path of the walking amputee. Based on the kinematic analysis, the kinetic data can be transformed into coordinate systems relative to specific segments or joints of the limb. Surface electrodes affixed to the skin above important muscle groups can pick up signals of gross muscle activity. This last component of information can be potentially very valuable, because it can help distinguish ineffective muscular performance from a misaligned prosthesis. There is room for affordable and portable technologies to enhance the research tools described with alternatives better suited to the clinical environment for ease-of setup, use, and interpretation. Lightweight load cells that can be positioned within a prosthesis have been developed, as have versatile data capture systems, allowing data collection over many sequential steps and in nonlaboratory environments (different surface terrains, different inclinations). Software to interpret and apply such data to fitting (e.g., suggest appropriate alignments), however, is only in its formative stages of development.

698

ARTIFICIAL LIMBS

The design of the hip joint has remained essentially the same through most of the latter half of this century, in part, because of the low demand for it. At this amputational level the energy requirement for an amputee to walk is so great that a high level of functionality is not commonly restored.

UPPER LIMB PROSTHETICS Amputation of the upper limb accounts for an estimated 12,000 surgical procedures a year in the United States and 15% to 20% of all amputations of major extremities. The most prevalent cause of upper extremity loss is trauma. Over 90% of all upper limb amputations result from major tissue damage from fractures, burns (electrical, chemical, or thermal), frostbite, and machinery accidents (11). Amputation of the upper limb often has a long-term impact on the patient’s life. Although lower limb amputations occur primarily in elderly dysvascular persons, the majority of upper limb amputations occur in young men between 20 and 40 years of age. Further, the upper limb has increased anatomical complexity and a need for finer control, making effective prosthetic design more challenging. Because of the low benefit-to-effort ratio of upper limb prosthetics, more than 50% of all upper limb amputees choose to forego prosthetics of any kind (12). Reasons for rejecting the prosthetic device include effective adaptation to a life with one hand, poor training or lack of skill in using the prosthetic device, and public perception of the prosthetic device. For those amputees who elect to use a prosthetic device, the choice of components varies considerably based on the level of amputation, the type of action required, the size and strength of the residual limb, and the desire to use body-controlled or electric components. Amputational Surgery The percentages of amputees who use prostheses vary by amputational level: partial hand/wrist disarticulation (12%), transradial (57%), elbow disarticulation (3%), transhumeral (23%), and shoulder disarticulation (5%). The level of amputation is often chosen to retain as much limb length as possible while ensuring effective wound healing. Partial hand amputations (including the loss of digits) account for the majority of upper limb amputations. An important aspect of partial hand amputation is to retain the functional capability of the thumb, providing apposition so as to retain grasping capability. Wrist flexion and extension are also maintained, as are forearm pronation and supination, functions which can then be transferred to the prosthesis. Adequate palmar skin must be available for this level of amputation, as the tactile palmar skin is used to cover the stump and to provide the grasping surface. Wrist disarticulation is typically performed when the thumb or fingers are lost. It is also performed as secondary surgery after partial hand amputation when functional apposition is not achieved. Wrist disarticulation offers the advantages of allowing full forearm pronation and supination, retaining an oval or rectangular residual limb that is a better holding shape in a socket and a more load-tolerant distal end because the distal ends of the radius and ulna are more loadtolerant surfaces. However, it can be difficult to fit a pros-

thetic device to the longer limb, thus a surgeon might choose a transradial amputation instead. Unlike the corresponding transtibial amputation, in transradial amputation both bones in the residual limb are transected to the same level to provide the largest amount of pronation/supination possible. As the length of the limb decreases, so too does the available degree of pronation/supination. Thus a long residual limb is preferred. Even if the residuum of the forearm is too small to attach an adequate transradial prosthesis, retention of the elbow is still highly desirable as a power source for body-powered cables. Although loss of a single upper limb is damaging, the loss of the second can be devastating. The sense of touch and proprioception provided by the hand is vitally important in daily living. The Krukenberg procedure provides an alternative for bilateral upper limb amputees (13). In this procedure, the radius, ulna, and associated muscles are separated, creating a prehensile organ capable of sensation and the capability to grasp objects. Because of the unsightly appearance of this technique, it is often reserved for bilateral amputees, blind amputees, and amputees in foreign countries where modern prosthetic devices are unavailable. An elbow disarticulation provides many of the same advantages as a knee disarticulation in lower limb amputees. The flared end of the residual limb at the condyles of the humerus provides suspension of the prosthetic socket and allows transmission of humeral rotation to the prosthetic device, alleviating the need for a separate component. The longer limb length is more advantageous as a moment arm and, like the wrist, the residual end provides a more load-tolerant residual limb. However, it can be difficult to fit a prosthesis, and the standard body-operated prosthesis for this type of residuum requires external locking hinges which often damage the amputee’s clothing. The long transhumeral amputation provides good power for body-controlled systems. Additionally, this level provides excellent sites for obtaining the signals used with a myoelectric terminal device, which controls an active device based on neural signals measured with surface or implanted electrodes, and provides the prosthetist with more options in location and access to components than the short transhumeral amputation. The short transhumeral amputation still provides adequate control of body-powered devices but lacks the appropriate power often needed to run such devices. Transhumeral amputations shorter than 30% of the original length may be considered shoulder disarticulations for all practical purposes. The shape of the shoulder is maintained by the soft tissue, providing a more natural looking contour. The amputee, however, has little to no excursion available for flexion, extension, or abduction of the prosthesis. Thus the use of a prosthesis is difficult. Prosthetic Design and Fitting Unlike lower limb sockets, the upper limb socket is not a weight-bearing device. Despite this difference, the need for a close yet comfortable fit is necessary for proper fitting. During the fitting process, the prosthetist pays close attention to the distribution of pressure, especially over bony areas such as the epicondyles and olecranon. The socket design consists of a double wall, the first to provide contact with the limb, and the second to provide an outer, cosmetic, stable shell.

ARTIFICIAL LIMBS

The distal area of the socket must provide adequate relief so that lifting with the forearm does not cause extreme pressure on the soft tissue between the radius and the socket wall. If the terminal device is body-powered, a sock should be worn on top of the residual limb to help reduce the pressure at the edges of the socket. However, myoelectric terminal devices require contact with the skin directly. Hence a sock cannot be worn. To avoid rubbing and restriction of shoulder motion, transhumeral sockets are generally fitted to those patients with between 30% and 90% of the humerus remaining. Using such a device, forearm flexion to 90% and abduction to 30% should be possible (14). The components of the prosthetic system include the terminal device, wrist unit, elbow unit, shoulder harness, and control system. Because of the large number of manufacturers and components, special attention must be given to the compatibility of all devices in the system, particularly electric components that may have special requirements. Terminal Devices. Terminal devices can be divided in two categories, passive and prehensile (or active). The prehensile device is also divided into two categories, body-powered and externally powered. The passive terminal device is the most prescribed form of terminal device. The passive hand is often simply used as a cosmetic device, although many passive devices have been designed for recreational activities, such as fishing, bowling, baseball, archery, and golf, though they require no motion or are passively positioned by the amputee. A passive hand is formed from a lightweight foam with central wires in the fingers covered with a cosmetic glove made from polyvinyl chloride or silicone rubber to provide a skinlike appearance. In a body-powered prosthesis, the amputee provides all of the energy to run the device. The body-powered terminal device is often designated by the opening or closing method. Generally, the device is characterized as either voluntary opening (VO) or voluntary closing (VC). In each case, the amputee applies a force by way of a control cable to provide the voluntary motion, and a spring or rubber band acts to return the device to its rest state. The force applied in grasping an object by the terminal device is determined by the spring force of the closing device for the VO and by the strength of the amputee in the VC. Such systems have the advantage of offering the patient sensory information based on the degree of motion of the controlling harness and shoulder position. The body-powered prehensile device can be one of two forms, a hook or hand. Quick release couplings allow easy interchange between them. The hook provides a more functional advantage as it provides superior prehension and visual feedback to close the control loop, especially when grasping small objects. The hook is most often designed in a split fashion where one finger is stationary and the other is driven by the control cable to provide lateral prehension. The hook prosthesis is also less expensive, more reliable, and far more sturdy than its hand counterpart. The body-powered hand device offers the advantage of cosmetic appearance over the rather obvious hook prosthesis. Rather than providing lateral prehension like the hook, the hand provides a palmar grasp, usually with a pinching motion between the thumb and the first and second fingers. Because of the increased complexity of the hand terminal device,

699

friction in the joints is high and the overall efficiency (and pinch force) is low. Other disadvantages include limited functionality, reduced reliability, and ease of damage to the plastic glove often worn over the hand. Externally or electric-powered prehensile devices use motors to bring opposing surfaces into contact to grasp an object. Currently no hand devices exist which offer the independent action of each finger in the device, although an electric hook device is available with independent motors to control each hook for greater flexibility and power. The externally powered prosthesis offers the advantage of improved appearance, improved output-to-input force ratio, range of motion, and a lesser degree of harnessing than a body-powered prosthesis. Like body-powered devices, their components are usually interchangeable, depending on the associated task. The electric devices typically run on a 6 V motor, providing grip strengths as high as 120 N (15) and closing times as fast as 0.8 s (16). Although growing in popularity, these devices have some disadvantages. The electric devices are more expensive and complicated than body-powered devices, necessitating continual maintenance. Also, such devices operate more slowly than similar body-powered devices. In recent studies, patients using both forms of terminal device took about twice as long to perform similar tasks with the electric device (17). The cosmetic glove typically worn over the electric prosthesis, like the body-powered equivalent, is often damaged but also often hinders operation of the electric device. Wrist. Commercially available prosthetic wrist units provide two functions: attachment of a terminal device and positioning for the terminal device. Wrist units can augment an amputee’s ability to supinate/pronate the forearm by rotating the terminal device. Wrist units are generally only passive devices, using friction to hold the terminal device in place. Any adjustment to the position must be accomplished by prepositioning the device with the opposing hand. The units are easily locked in one position, though mechanisms to allow for quick release and change of position are available. Electric wrist rotators are beginning to be used in more advanced systems. These units offer additional independent control but suffer from added weight and low torque. Elbow. Both body-powered and electric-powered elbows are available (Fig. 5). Although most body-powered, transradial sockets require harnessing, self-suspension sockets provide freedom from harnessing or cables when used with electric terminal devices. Such a system is particularly useful with short or very short transradial amputations and an electric terminal device. The elbow unit simulates rotation at the elbow through a turntable device located in the transhumeral prosthesis and allows flexion through hinges mounted internally or externally to the lateral sides of the prosthesis. The force for arm flexion derived from a cable system is similar to the voluntary opening terminal device. The two systems are often combined in a single tension line to provide flexion of the elbow when the elbow is unlocked and opening of the terminal device when the elbow is locked. Like the wrist unit, this device offers several locking positions (usually seven to eleven discrete positions of flexion). The elbow forearm lift-assisted units are often friction-held or spring-balanced to remove some of the weight of the forearm from the residual limb and reduce the

700

ARTIFICIAL LIMBS

ing scapular motion. Although it provides a relatively large amount of force, it requires an anchor point, usually at the waist, to accomplish the task. This motion is often used for locking and unlocking elbow units in transhumeral or higher levels of amputations. Other actions, such as scapular abduction and chest expansion, are often used to accomplish these tasks if other motions are unavailable. The harness system is dictated by the level of amputation. Transradial amputees most often use the figure-eight harness [Fig. 5(upper)], whereas transhumeral amputees require a more complex harness with an additional support strap. The transhumeral harness uses one strap to keep the prosthesis suspended from the residual limb and a single control cable to operate the terminal device. The transhumeral harness utilizes an elbow-lock cable strap to operate the elbow unit in much the same manner as the figure-eight harness used for the terminal device. Myoelectric Control Systems

Figure 5. Upper limb prostheses. (upper) Figure eight method for harnessing an upper extremity transradial prosthesis (From N. Berger, Upper limb prosthetic systems, in Atlas of Limb Prosthetics, Surgical and Prosthetic Principles, American Academy of Orthopaedic Surgeons, St. Louis, MO: C. V. Mosby Company, 1981, with permission.) (lower) A NY electric elbow with prehensile actuator (From Hosmer Dorrance Corporation, The NY elbow system, Electric Components, 1997, with permission.)

force necessary for flexion. If more than 90% of the humerus remains, then internally locking elbows are not possible, and externally mounted hinges must be used. Electric-powered elbows incorporate either myoelectric signals or switch controls and are often used in conjunction with myoelectric devices. Because of high cost, high weight, and low reliability, electric elbows are rarely used, and instead a hybrid system is used consisting of a myoelectric terminal device and a standard body-driven elbow unit to provide separation of control. Shoulder Harness. To provide power for the body-controlled components, a control harness is often used. One of the best methods for providing motion is glenohumeral (shoulder) flexion, provided enough humeral length remains. Between approximately 180 N and 270 N of force is generated by the average adult in shoulder flexion (17), providing adequate power for flexing elbows and/or opening terminal devices. Another method of providing force for operating the prosthetic control system is shoulder elevation and depression, captur-

Although not a new technology, myoelectric control in prosthetic systems, has been at the forefront of research for many years. Myoelectric control is a relatively simple concept. When a neural signal is transmitted from the brain via the spinal cord, which in an intact individual would produce muscle contraction, this signal can be detected by electrodes inserted under the skin or applied to the skin surface. These signals can be used to control specific devices. The myoelectric system consists of five main components, the signal source (muscle activation), the electrodes, the controller, the power source (battery), and the prosthetic device. Analysis of the electromyographic (EMG) waveform demonstrates that the majority of the energy in the signal lies in the 30 to 300 Hz frequency range and that the maximum peak-topeak amplitude of the signal ranges from a few microvolts to several millivolts (18). Correlation between the activation and the properties of the signal can be determined and a method of control obtained. It is important to note that this method of control is suitable for atrophied, partially innervated, and remnants of muscles often seen in the residual limb. The surface electrodes, typically made of gold or stainless steel with a surface area of less than 1 cm2 or less, provide one of the most problematic areas in myoelectric systems. The skin is a natural electrical insulator and often distorts the myoelectric signal. Future areas of research may include percutaneous conductors, myoacoustic receptors, or implanted telemetry systems to overcome this problem. Another common difficulty is that relative movement between the electrode and skin creates noise that is often greater than the myoelectric signal. To alleviate this problem, the electrodes are set at a fixed distance (2 to 3 mm) from the skin surface with a conductive cream or gel filling the intermittent gap. However, the amplitude of the myoelectric signal degrades rapidy as the distance between the source and the electrode increases. Further, at these greater distances, as other local active muscles contract, they produce crosstalk in the signal, causing erroneous control. The myoelectric controller controls the power to the motor. It typically consist of three separate components, an amplifier, a signal processor, and a logic unit. An adjustable amplifier increases the myoelectric signal amplitude to a suitable

ARTIFICIAL LIMBS

701

Figure 6. Myoelectric signal (EMG) processing in a typical myoelectric control system (From D.S. Childress, Control of limb prostheses, in Atlas of Limb Prosthetics, American Academy of Orthopaedic Surgeons, St. Louis, MO: Mosby Yearbook, Inc., 1992, with permission.)

level (gain of typically 10,000 to 100,000). A differential amplifier, including a common electrode and two active electrodes for each channel, is most often used to accentuate the myoelectric signal interpreted by the logic unit and also to reduce amplification of noise. A signal processor overcomes noise inherent in the myoelectric signal and sends a more meaningful control waveform (19) (Fig. 6). Typically, the mean absolute value of the signal is used. Tests have, however, shown that the amount of myoelectric signal becomes more accurate as the average signal is sampled over a longer time, creating a delay. A typical delay of 0.2 s is used in modern myoelectric prosthetic control systems. The battery provides power to the motor for the terminal device (or electrical component). Most electric prosthetic devices are powered by secondary cell (rechargeable) batteries, such as nickel-cadmium, although some utilize primary cell (nonrechargeable) batteries when necessary. The battery typically provides between 6 V and 12 V to the motor. Although the majority of the electric systems operate at 6 V, some systems are designed to operate at multiple voltage levels in case primary cell batteries are needed to run the unit. Because one prosthetic device rarely meets the needs of any amputee, electronic prosthetic systems are designed to be modular, providing the amputee with the ability to interchange components based on desire or need. In theory, almost any electric device can be controlled through a myoelectric signal, provided the voltage requirements of each device are consistent with the other components in the system. CURRENT RESEARCH Interfacial Mechanics In recent years much research effort in prosthetics has concentrated on better understanding how mechanical stresses are distributed at the interface of a residual limb and prosthetic socket, particularly in lower limb prosthetic applications which involve high load bearing. A better understanding of interfacial mechanics and how design features of the residual limb and prosthesis affect them will help prosthetists create artificial limbs that reduce the risk of skin breakdown while maintaining stability.

A number of pressure-sensing instruments and a limited number of pressure/shear stress measurement devices have been used to quantify interfacial stresses (20). Such measurements are difficult to acquire because of sensor size and mass restrictions in the confined interfacial environment and a need to limit alterations of the natural interface. Nevertheless, some insight has been achieved and interfacial stress sensitivity to different parameters has been assessed. An interesting finding consistent with clinical experience is that diurnal and long-term changes in residual limb shape and/or material properties cause substantial changes in interfacial stress distributions. Residual limb shape changes are an important challenge for prosthesis users, and establishing relationships between interfacial stresses and shape change is an important goal for researchers interested in interfacial mechanics. One tool with potential to enhance understanding of the effects of residual limb shape changes and other features on interfacial mechanics that does not suffer from the measurement problems of interfacial stress transducers is scientific computing. At present, computer-aided methods of fabricating sockets are largely limited to measuring the residual limb shape and carving the rectified positive replica. The application of engineering design principles based on modeling the geometry and material behavior under anticipated loading could considerably enhance socket rectification. Modeling the interaction between the prosthetic socket and the residual limb is currently being carried out in research laboratories. The three inputs required for such models are (1) the geometry of the residual limb and prosthetic socket, including the individual geometries of the bones, the soft tissues, the socket liner, and the socket shell; (2) the material characterization of each geometric component; and (3) the external dynamic load experienced by the socket, plus other external constraints, such as suspension, friction, and suction. The reference against which the predictive ability of the models is assessed is the experimentally measured contact stresses between the prosthetic socket and residual limb. Finite element methods of structural analysis are commonly used in the models because of the complex geometries involved and the very nonlinear behavior of the component materials, especially the soft

702

ARTIFICIAL LIMBS

Figure 7. Finite element model prediction of resultant shear stress magnitudes (in MPa) on the surface of a residual limb. An axial load equal to body weight was applied and homogeneous, linear, isotropic material properties were assumed for all materials.

tissues (Fig. 7). Finite element models potentially can predict interfacial pressures and shear stresses for proposed socket designs, information that could then be used to optimize socket design features. A number of hurdles must be crossed to achieve such a clinical goal. At present magnetic resonance imaging and computed tomography offer ways to visualize the subsurface structures. Both of these are cumbersome and expensive. Some investigators have shown that ultrasound can be used to build a picture of the residual limb’s internal and external geometries (21). Such a technology may be viable within a prosthetic production facility. Electromechanical indentation devices to assess the soft tissue material behavior have been developed in many research labs and have also been used in arrays for wheelchair cushion design. The most challenging aspect of the design is predicting the stress pattern resulting from a particular design and activity level. Finite element analysis for nonlinear material properties and interfacial frictional behavior is still very much in its early development. Further, current finite element models treat all the muscles as though they were a uniform material when in fact muscles are fibers contained within specific sheaths and have corrections to the skeletal system at specific places; in addition, they may also stiffen when they are active. These features need to be taken into account in the models. Though software tools exist for finite element analyses, the computational cost limits their use to the most high-end engineering workstations and supercomputers. None of these challenges, however, is insurmountable.

Advanced Upper Limb Control Methods Much of the advanced work in upper limb prosthetics is governed by problems associated with myoelectric prostheses. Although such devices provide a more cosmetic look, improved function, and freedom from harnessing, they still suffer from substantial drawbacks. Two important issues include the need for multifunctional control systems and electronic and sensory feedback in the prosthesis. Multifunctional Control Systems. The standard myoelectric control system is characterized by the number of host sites required for the electrodes and the number of control states available to the muscle. Currently, both single and dual sites are available per action (i.e., flexion/extension or open/close). The one-site, one-state control [Fig. 8(a)] offers control of the device through electrodes placed on a single muscle. Activation of the muscle activates the device (usually open), and then springs or rubber bands return the device to the rest position. Two-site, one-state [Fig. 8(b)] control uses two such devices to control both the open and close motion of the device. Activation of one opens the device and activation of the other closes it. This method of myoelectric control is used most often. The one-site, three-state control (Fig. 8(c) allows the user to change control between open and close based on the magnitude of the myoelectric signal. This method uses two threshold levels to determine the close and open states. A time delay in the control allows the user to transfer directly from off to close if so desired.

ARTIFICIAL LIMBS

703

contact. Such a system offers more control than a conventional myoelectric controller but requires more gradients in the EMG signal than typical two-site, two-state devices (22).

Figure 8. Typical control options for myoelectric prosthetic systems. (a) one-site, one-state; (b) two-site, two-state; (c) one-site, three-state.

Advanced control systems now under development utilize two muscles in much the same method as the two-site, twostate control but with a greater degree of multifunctional motion. Proportional control or procontrol is used in some systems to allow the rate and magnitude of the myoelectric signal to provide additional movement. In this system, the rate of motion of the device (elbow, wrist, or terminal device) is proportional to the magnitude of the EMG signal rather than the binary (threshold) on-off/constant-velocity motion of other devices. The harder the control muscles are flexed, the faster the motion of the device. Additionally, the rate of muscle activation provides additional multistate control. Holding the arm in a single position causes the elbow to lock in position, switching control to the hand or wrist. A quick coactivation of both control muscles allows the user to return to the elbowcontrol mode, thereby affording a full range of motion to the amputee with limited control sites. Other systems use enhanced feedback features to provide improved control over each component. A vibration sensor and appropriately designed controller is used in some systems to help control the grasp of the terminal device. This multistate control allows the user to control both the position and force of the hand with two muscles. In this case, the device is a voluntary opening hand, where tension in the extensor opens the hand proportionally to the magnitude of EMG signal. When a sensor in the hand comes in contact with the object, movement is stopped. Tension in the flexor muscles activates the HOLD state and an automatic force grip is activated. The automatic force control detects movement of the object from the sensor and adjusts to prevent slippage. While in HOLD mode, the user can override the control in a SQUEEZE mode to apply additional force to the object or can initiate a RELEASE mode to return to the original position of

Electronic and Sensory Feedback. The two types of feedback important in prosthetic control are electronic feedback, the feedback provided to the electronic control system, and sensory feedback, the feedback provided directly to the amputee. Electronic feedback provides feedback to enhance the function of the prosthetic device, such as position, joint angle, joint torque, and velocity. Such measurements are provided by force and angle transducers in the device. Such feedback ultimately enhances control of the prosthetic device. Sensory feedback or proprioception is critically valuable to the amputee. Without such feedback, the terminal device may provide too much or too little force, thereby damaging or dropping a grasped object. In the past, researchers have used inflatable bags to provide pressure on the residual limb, which corresponds to the grip force of the terminal device. The most popular of the current methods of proprioception includes electrical stimulation and vibration. One problem with such methods is the limited amount of feedback provided. Most patients can differentiate among only approximately five levels of stimulus, limiting the value of the feedback information transferred in this manner. Additionally, the feedback provided by electrical stimulation can cause interference or crosstalk with the sensitive myoelectric system. Further advances in sensory perception are possible using direct communication with peripheral nerves (23), but research is still continuing in this area.

BIBLIOGRAPHY 1. A. Selincourt (ed.), Herodotus, the Histories, New York: Penguin Books, 1954. 2. M. Vitali et al., Amputations and Prosthetics, 2nd ed., London: Bailliere Tindall, 1986. 3. I. M. Rutkow and P. H. Marlboro, Orthopaedic operations in the United States, 1979 through 1983, J. Bone Joint Surg., 68-A 716– 719, 1986. 4. E. M. Burgess, Major amputations, in P. F. Nora (ed.), Operative Surgery—Principles and Techniques, Philadelphia: Lea and Febiger, 1972. 5. E. M. Burgess, Below knee amputation, Surg. Techniques Illustrated, 3: 59–67, 1978. 6. D. Dean and C. G. Saunders, A software package for design and manufacture of prosthetic sockets for transtibial amputees, IEEE Trans. Biomed. Eng., BME-32: 257–262, 1985. 7. C. D. Fillauer, C. H. Pritam, and K. D. Fillauer, Evolution and development of the silicone suction socket (3S) for below-knee prostheses, J. Prosthet. Orthot., 1: 92–103, 1989. 8. J. W. Michael, Prosthetic knee mechanisms, Physical Medicine Rehabil.—State Art Rev., 8: 147–164, 1994. 9. E. M. Burgess et al., The Seattle prosthetic foot—a design for active sports: Preliminary studies, Orthot. Prosthet., 37: 25–32, 1983. 10. D. Rowell and R. W. Mann, Human movement analysis, Soma, 3: 13–20, 1989. 11. T. J. Moore, Amputations of the upper extremities, in M. W. Chapman (ed.), Operative Orthopedics, Philadelphia: J. B. Lippincott Company, 1993.

704

ASSISTIVE DEVICES FOR MOTOR DISABILITIES

12. A. L. Muilenburg and M. A. LeBlanc, Body-powered upper-limb components, in D. J. Atkins and R. H. Meier III (eds.), Comprehensive Management of the Upper Limb Amputee, New York: Springer-Verlag, 1989. 13. S. K. Jain, The Krukenberg operation, in G. Murdoch and A. B. Wilson (eds.), Amputation: Surgical Practice and Patient Management, Oxford: Butterworth Heinemann, 1996. 14. D. D. Kilmer and J. S. Lieberman, Principles in prosthetic fitting. In M. W. Chapman (ed.), Operative Orthopaedics, Philadelphia: J. B. Lippincott, 1993. 15. C. W. Heckathorne, Components for adult externally powered systems, in Atlas of Limb Prosthetics, American Academy of Orthopaedic Surgeons, St. Louis, MO: Mosby Yearbook, 1992, pp. 151–173. 16. W. F. Sauter, Electric pediatric and adult prosthetic components, in D. J. Atkins and R. H. Meier III (eds.), Comprehensive Management of the Upper Limb Amputee, New York: Springer-Verlag, 1989. 17. D. G. Shurr and T. M. Cook, Prosthetics and Orthotics, Norwalk: Appleton & Lange, 1990. 18. R. N. Scott, Biomedical engineering in upper-extremity prosthetics, in D. J. Atkins and R. H. Meier III (eds.), Comprehensive Management of the Upper Limb Amputee, New York: SpringerVerlag, 1989. 19. D. S. Childress, Control of limb prostheses, in Atlas of Limb Prosthetics, American Academy of Orthopaedic Surgeons, St. Louis, MO: Mosby Yearbook, 1992, pp. 175–198. 20. J. E. Sanders, Interface mechanics in external prosthetics: Review of interface stress measurement techniques. Med. Biol. Eng. Comput., 33: 509–516, 1995. 21. P. He et al., A PC-based ultrasonic data acquisition system for computer-aided prosthetic socket design, IEEE Trans. Rehabil. Eng., 4: 114–119, 1996. 22. P. J. Kyberd et al., A clinical experience with a hierarchically controlled myoelectric hand prosthesis with vibro-tactile feedback, Prosthet. Orthot. Int., 18: 56–64, 1993. 23. R. N. Scott, Feedback in myoelectric prostheses, Clin. Orthop. Relat. Res., 256: 58–63, 1990. Reading List S. G. Zachariah and J. E. Sanders, Interface mechanics in lower-limb external prosthetics: A review of finite element model, IEEE Trans. Rehabil. Eng., 4: 288–302, 1996. American Academy of Orthopaedic Surgeons Staff, Atlas of Limb Prosthetics: Surgical, Prosthetic, and Rehabilitation Principles, 2nd ed., St. Louis, MO: Mosby Yearbook, 1992. W. H. Bohme, Atlas of Amputation Surgery. New York: Thieme Medical Publishers, 1987. P. A. C. Lim, Advances in prosthetics: A clinical perspective, in T. N. Monga and K. P. Zimmerman (eds.), Physical Medicine Rehabil. Clinics: State Art Rev.—Advances in Rehabil. Technol., 11: 13– 38, 1991. J. Perry, Gait Analysis, Normal and Pathological Function. Thorofare, NJ: SLACK Incorporated, 1992. E. M. Burgess, R. L. Romano, and J. H. Zettl, The Management of Lower Extremity Amputations, Prosthetic and Sensory Aids Service, V.A., Washington, DC, 1969. G. T. Sanders, B. J. May, and J. Milani, Lower Limb Amputations: A Guide to Rehabilitation. Philadelphia: R. A. Davis, 1986. Lower Limb Prosthetics, Prosthetic Orthotic Publication, New York: New York University Press, 1990. M. B. Silver-Thorn, J. W. Steege, and D. S. Childress, A review of prosthetic interface stress investigations, J. Rehabil. Res. Develop., 33: 253–266, 1996.

C. W. Radcliffe, Locomotion and lower limb prosthetics, Bull. Prosthet. Res., 16: 167–187, 1974.

J. E. SANDERS S. G. ZACHARIAH B. J. HAFNER University of Washington

ARTIFICIAL NEURAL NETWORKS. See NEURAL ARCHITECTURE IN

3-D; NEURAL NETS BASED ON BIOLOGY.

ASSEMBLERS FOR PROGRAMS. See PROGRAM ASSEMBLERS.

ASSEMBLY LANGUAGES. See INSTRUCTION SETS. ASSEMBLY, SURFACE MOUNT. See SURFACE MOUNT TECHNOLOGY.


}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP


Wiley Encyclopedia of Electrical and Electronics Engineering Assistive Devices For Motor Disabilities Standard Article Vijay Kumar1, Tariq Rahman2, Venkat Krovi3 1University of Pennsylvania, Philadelphia, PA 2Alfred I. duPont Institute and the University of Delaware, Wilmington, DE 3University of Pennsylvania, Philadelphia, PA Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6603 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (403K)

●

●

● ●




Abstract The sections in this article are Prosthetics Assistive Robot Manipulators Wheelchairs Teletheses and Human Extenders: Research Issues Design and Manufacture Conclusion About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.


❍

❍ ❍


704


ASSISTIVE DEVICES FOR MOTOR DISABILITIES There are many examples of assistive devices for people with manipulative and locomotive disabilities. These devices allow disabled people to perform many activities of daily living, thus improving their quality of life. People with disabilities are increasingly able to lead independent lives and play a more productive role in society. In the case of children with disabilities, such assistive devices have been shown to be critical to their cognitive, physical, and social development (1). The earliest assistive devices were prosthetic limbs, dating back to 500 B.C. (2). The early wheelchairs, in contrast, found widespread use less than 300 years ago. These simple prosthetic limbs and wheelchairs have since evolved into more complex multi-degree-of-freedom mechanical and electromechanical devices. In particular, robotic technology has been used to enhance the quality of life of people with disabilities, primarily by enhancing a person’s capability for independent living and vocational productivity. An assistive robot (also called a rehabilitation robot) may be viewed as distinct from a prosthesis in that it is not attached to the user, but may reside on a table top, or on the side of a wheelchair, or on an independent mobile base. However, this distinction may blur in the case of electromechanical aids that are worn by the user. The goal of this article is to review the state of the art in the technology for assistive devices for people with disabilities, with a particular focus on the technology that is loosely referred to as robotics. This includes articulated orthoses as well as robotic devices. In the process, we review research that has been done by us and by other groups on assistive devices for manipulation and locomotion. We are less interested in examples of devices that simply perform the mechanical function of a person’s limb and instead focus on assistive aids that have broader applications. Further therapeutic applications are beyond the scope of this article. Similarly, orthoses that strengthen limbs and spines, or prevent deformities are not considered here. Instead the main goal is to provide the reader with an understanding of how the technology and science that underlies robotics can be used to develop assistive devices for people with manipulative and locomotive disabilities. J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.


PROSTHETICS The basic goal of a prosthetic device is to provide to a person with a disability an aid that can perform the function of one or more limbs. We focus on upper-limb prosthetics for people with manipulative disabilities. The body-operated Bowden cable arm came into widespread use after World War II and still remains the prosthesis of choice for many amputees, primarily because of its inherent kinesthetic feedback associated with cable control (3,4). However, with the advent of new technologies such as the transistor and the microprocessor, externally powered devices that augment human strength became more prominent. There are two main approaches to controlling such externally powered devices: 1. Activation of prosthetic joints with the aid of myoelectric signals from intact musculature 2. Control by displacement signals obtained from body movements Electromyographic Control Electromyographic (EMG), or myoelectric, control uses the electric signal due to depolarization of the cell membrane of muscle fibers during contraction (5). The signal is sensed through electrodes, amplified, processed, and then used as input to the actuators. It was first used in prosthetics by Reiter in the early 1940s (4). Later, Bottomley (6) used EMG signals for the proportional control of prehension in the English hand. EMG control has been employed in many prosthetic arms with limited degrees of freedom. The main drawback of EMG control (6–8) is its essentially open loop character due to the absence of position proprioception. Use of EMG for multiaxis prostheses or robots is deemed inferior because controlling hand position in space by individual joint velocity control is considered mentally taxing (9). Although alternatives such as resolved motion rate control (10) or end-point control (8) can in principle overcome this shortcoming, it is very difficult to draw the control signals from the muscles that are directly related to the movement, which is an important requirement for natural control of the arm. A variant on the EMG control is to use impedance properties of muscles, as opposed to the more traditional velocitycontrolled EMG (11). Although this approach appears to have its advantages, it still fails to provide proprioception to the user and the mode of operation is still open-loop. There have been many attempts to develop an artificial sensory system for proprioception (12). Although artificial exteroception does provide cues of position and force, it is a poor substitute for proprioception. There is a body of research describing the application of pattern-recognition techniques in the control of myoelectrical controlled prostheses. Because synergistic muscle groups are responsible for activating the joints of the natural limb, any seemingly natural control scheme must take as input various EMG signals from the shoulder and chest. The key technical challenge is to interpret the patterns of EMG and to match these patterns to specific movements (13). However, myoelectric signals are often inconsistent and the reliability and the benefits of such an approach is questionable.

705

Body Movement Control Visual and auditory feedback are slower, less automated, and less programmed than the normal proprioceptive feedback (9). With EMG control, the amputee relies on vision and exteroceptive feedback to determine how well his intentions have been executed by the prosthesis. In contrast, if the movement of the prosthesis is physically linked to body movement, not only does it reduce demands on system response, but it also provides proprioceptive feedback, and hence, a closed-loop system. Body movement control can be divided into discrete and continuous inputs. Discrete signals that are effected by body movements include switches operated by digits in phocomelic children (14), shoulder displacement switches (15), feet and shoulder switches (16), and various body switches (4). The disadvantage of discrete control is that it usually relates the duration of switch closure to the distance moved. This relationship does not conform to natural modes of control, and coordination of multijoint movement becomes difficult. A continuous signal, as opposed to a discrete signal, offers superior control. For example, a person who uses the biscapular movement of the shoulders to flex and extend the elbow of a conventional cable-operated arm has a sense of being linked to the arm. The user exerts a force on the cable which moves the artificial joint. The movement at the joint is linked to the amount of movement permitted at the shoulder. The prosthesis therefore acts as an extension of the user and provides force and position information to the user. This exchange of information and energy signals is termed bilateral control and is an important feature of telerobotic systems (17). Bilateral control, and in particular impedance control as described by Hogan (18) share much in common with powered prosthesis control using body movements. Simpson (19) attempted to realize this basic idea by developing the Edinburgh arm, a five-degree-of-freedom pneumatic arm prosthesis for amelic and phocomelic children victimized by thalidomide. The control of the prosthesis was performed by movements of the shoulders which were conveyed via position servo systems of the joints. It was claimed that the advantage of this approach lay in the full position awareness of natural body movements provided by joint proprioceptors. In this case, the major responsible joint was the sternoclavicular joint. Since movements were conveyed to the hand, and the prosthesis served as an extension of a joint, this control concept has been referred to as extended physiological proprioception (EPP). This basic idea has been pursued by (20–22) and it has been demonstrated to be more intuitive to use and superior in tracking tasks (23,24). The main difficulty in EPP based systems is the coordinate control of multiple joints. O’Riain and Gibbons (21) investigated proprioceptive control by a microprocessor using shoulder movement. This system employed a repertoire of input/ output linkages (relationships). These linkages were designed to overcome limitations in functionality. However, this gain is accompanied by decreased position awareness. Other devices of note are the Vaduz hand (6), developed in the 1950s, which used muscle bulge to operate a switch-activated position servomechanism to control an electric prehension device. This system contains concepts of EPP in that it provides awareness of force. For a review of powered limb prosthetics see Ref. 4 and for feedback aspects of prosthetics see Ref. 3.

706


ASSISTIVE ROBOT MANIPULATORS Early attempts at developing rehabilation robots included the Rancho ‘‘Golden’’ arm (25), the Heidelberg arm (26), the VAPC arm (27), and the Johns Hopkins arm (28). Although these devices saw limited use by consumers, they established the foundation for further development in the field. There are a number of rehabilation robots currently available or in development. The most well-known device is the MANUS (29), which is a wheelchair-mounted seven-axis (plus gripper) robot. The MANUS (manufactured by Exact Dynamics), a Dutch project, was designed from the start with the disabled person in mind. It was a unique collaboration between the engineering and rehabilitation worlds, which rendered a well-engineered, quiet, and aesthetic device. The MANUS folds up into an unobtrusive position at the side of the wheelchair and folds out when commanded. Its present inputs include a 16-button keypad, trackball, and joystick. The MANUS allows task space control. In other words, the user may directly control the motion of the end effector in Cartesian coordinates (translations along and rotations about Cartesian axes). This is in addition to the less sophisticated joint spacecontrol mode in which each joint is controlled independently. There are currently approximately fifty users of the MANUS, mainly in the Netherlands. Another project that has enjoyed relative success is the Handy I (Rehab Robotics Ltd., Staffordshire, UK) (30). The Handy I uses an inexpensive industrial robot arm (Cyber 310) to perform programmed tasks. The system is primarily used as a feeding device for children with cerebral palsy. The user uses a chin switch to activate the system and select the food through a scanning selector, which is then automatically brought up to the mouth. The system has enabled a number of children to feed themselves for the first time. Currently, over 80 of these systems are being used by disabled individuals in the United Kingdom. The RAID project (31) is a collaborative effort on the part of several European concerns. The aim of the project is to develop and demonstrate a prototype workstation for use by the disabled or elderly in a vocational setting. It consists of a sixdegree-of-freedom RTX robot placed on a linear track, and a structured workcell. A user may choose his or her preferred input device to control the robot by issuing high-level commands such as ‘‘pick up book three.’’ RAID is currently being evaluated in a number of rehab centers in Europe. In North America, there are three commercial projects of note. The first is DeVAR (desktop assistant robot for vocational support in office settings), a Palo Alto Veterans Administration/Stanford University collaboration that uses a PUMA-260 robot mounted on an overhead track that performs preprogrammed tasks in a highly structured environment (32). The DeVAR system has been evaluated by a number of individuals in various Veterans Administration centers, and notably by one highly motivated, disabled individual on a two-year trial in his work environment. The project yielded much information on cost/benefit and social issues, however a high price tag has prevented commercial success. The second project is the Robotic Assistive Appliance (RAA) developed at the Neil Squire Foundation in Vancouver, Canada (33). The RAA, which is the result of over 10 years of research in rehabilitation robotics, offers a human size manipulator at a workstation with six-degrees-of-freedom with

either programmed or direct control. The device is currently undergoing testing to assess its advantages over an attendant. The third commercial prototype is the Helping Hand (34) which was developed by Kinetic Rehabilitation Instruments (KRI) of Hanover, Massachusetts. The Helping Hand is a fivedegree-of-freedom modular arm, that can be mounted on either side of most powered wheelchairs. The arm comes with its own controller comprised of switches for the joint motors. It does not include a computer, which reduces cost and complexity. To date it has been evaluated in a number of Veterans Administration centers and has been approved by the Food and Drug Administration. However, it remains to be seen whether the Helping Hand will meet with long-term success. Even though the field of robotics has grown considerably in the last 20 years, from robots operating in the space shuttle to robots used to assist in surgery, there is a disappointing lack of progress in rehabilitation robotics. Rehabilitation robots have had limited success as commercial products because of the high cost, the poor interface between a complex electromechanical system and a human with limited capabilities, and the social stigma associated with a robot. Very often, the designer has a poor understanding of the needs of a disabled individual. The user often needs assistive devices that are customized to his or her needs and not necessarily a generalpurpose, complex rehabilitation robot.

WHEELCHAIRS Despite rapid scientific and technological progress in allied disciplines, there has been very little innovation in wheelchair design over the last 200 to 300 years. The folding wheelchair came in 1933, and powered wheelchairs were developed in the early 1970s (35). New materials such as plastics, fiberreinforced composites, and beryllium–aluminum alloys have found their way into the design and manufacture of lighter, stronger, and more reliable wheelchairs (36). The wheelchair industry has also benefitted from the development of lighter, efficient, durable, and reliable motors, better amplifiers and controllers, and most important of all superior batteries. There is considerable research and development activity focused on wheelchairs. Since the user is in intimate physical contact with the chair for extended periods of time, the contact surfaces, especially the seat, require a certain degree of customization to ensure comfort (37). Commercially available standup wheelchairs afford better seating and reaching, relief from pressure sores, and better health (38–40). They also allow users to operate equipment designed to be operated in a standing position (38). Conventional wheelchairs are difficult to maneuver in constrained spaces because they only have two degrees of freedom (forward/backward movement and steering). However, the Alexis Omnidirectional Wheelchair (41), TRANSROVR (42), and the European TIDE Initiative OMNI Wheelchair (43) can move omnidirectionally by adapting nonconventional wheels developed for use by robotic vehicles for this application (44,45). A number of computer-controlled wheelchairs have been developed in recent years, including the CALL Smart Chair (46), NavChair (47), TinMan (48), and WALKY (49). Wheel-


Figure 1. Photograph of the stair-climbing wheelchair rolling down stairs. (Courtesy, Professor Shigeo Hirose, Tokyo Institute of Technology)

chair systems with customized user interfaces, sensors, and controllers, suitably integrated (50), can potentially make the operation of a wheelchair much simpler and make it more accessible to people with disabilities. Such chairs may use a wide variety of sensors ranging from ultrasonic range sensors (51), cameras, encoders, accelerometers, and gyroscopes and any desired input device [communication aids, conventional joysticks, sip and puff switches, pressure pads, laser pointers, speech recognition systems, and force reflecting joysticks (52)]. Suitable control algorithms assist the user in avoiding obstacles, following features such as walls, planning collisionfree paths and traveling safely in cluttered environments with minimal user input (53–56). While motorized wheelchairs with sophisticated controls are well-suited to locomote on prepared surfaces, most are unable to surmount common obstacles like steps and curbs. Special purpose aids (57,58), including stairway lifts (59), stair climbers (60,61), and customized outdoor buggies have been developed for specific environments, but they are not versatile enough for multipurpose use. For example, a wheelchair that can go up and down any flight of stairs has remained an open research and development issue over the past couple of decades. One innovative proposal by Professor Shigeo Hirose (62) is shown in Fig. 1. A novel remote center mechanism (63)

707

moves the seat on an elliptical arc as the attitude of the chair changes and maintains the posture of the user independent of the wheelchair posture. A minimal degree of active control is required which is accomplished by a simple attitude sensor and a relatively small actuator. See Table 1 for a brief survey of available solutions. However, most of these solutions are not appropriate for unstructured outdoor terrains. Users cannot drive their chair on beaches, nor can they easily cross muddy patches and potholes. One approach to improving the mobility of a wheelchair by an order of magnitude involves the use of legs instead of wheels as locomotion elements. Advances in robotics have made it possible to build and control legged machines (64– 67). It is not difficult to imagine wheelchairs with legs climbing slopes, stepping over obstacles and walking on uneven terrain. A four-legged chair developed by the University of Illinois at Chicago and the Veterans Administration Hines Rehabilitation Research and Development Center based on research in quadruped walking (68,69) was developed in 1987. The walking chair was designed to enable the user to walk up and down stairs, steep slopes, across rough terrain, with curb weight less than 113.6 kg (250 lb) and capable of carrying a payload of 113.6 kg (250 lb). A full-scale prototype (70) design incorporating computer-controlled pantographic legs walked in the laboratory in October 1988, with a simple linear gait. However, it did not carry a passenger, and it was connected by a tether to a stationary controller. There are several inherent disadvantages in the concept of a legged chair. The legs are responsible for keeping the rider in a stable posture. There is a natural concern for safety that arises here. In wheeled systems, the wheels passively support the chair and do not require any sophisticated actuators or control electronics. In a legged system, stability must be maintained actively. Because of the complexity of the system, reliability is a natural concern. Furthermore, for stability, at least three support legs must be on the ground and a vertical line through the center of gravity must pass within the polygon formed by the support points. This implies that at least four legs are required to make a legged system walk—one leg is moved forward while three others support the chair. In the worst case, one leg must support half the weight of the chair and the user. This implies that each leg must have a strength- (payload-) to-weight ratio several times greater than one, with a payload of the order of a hundred pounds. The leg designs and actuators scale very poorly to such high payloads. Since the actuators must run off wheelchair batteries, and since there are severe restrictions on how large the chair can be [e.g., the maximum width must be less than 0.762 m (30 in.)], there are serious constraints that make it difficult to design a practical legged chair. An alternative design for a wheelchair for locomotion on uneven terrain tries to combine the advantages of legged locomotion (versatility, adaptability) with wheeled locomotion (reliability, superior stability) (71,72). One hybrid wheelchair has two powered rear wheels, two front castors, and two legs (72), as shown in Fig. 2. The experimental prototype is equipped with six dc motors, position and force sensors, and an on-board computer. It weighs 28.2 kg (62.0 lb) without the batteries and controller, and can climb a 1 ft curb with a payload of 68.2 kg (150 lb). The powered wheels are used to navigate on a flat surface as in a conventional wheelchair, while the legs and wheels are used to traverse uneven terrain. In

708


Table 1. A Survey of Available Methods (Technology) for Enhancing Mobility Solution

Advantages

Disadvantages

Architectural modifications (curb cuts, ramps, accessible elevators)

Usually low cost to consumers. Assists all ages and abilities. Often a simple technology with low maintenance. High consumer acceptance.

Transfer technologies

Can transfer to the vehicle most appropriate for the environment. Allow access to certain wheelchair inaccessible environments.

Regulations do not apply to private or historic buildings. Apply only in limited measures to apartment buildings. Many buildings do not comply with the law. Not applicable in most outdoor settings. May require assistance with a transfer.

Stair-climbing wheelchairs

Customized chairs (outdoor buggies) Curb climbers

Optimized for the environment. Low cost. For example, golf carts, outdoor chairs, and special purpose sand buggies.

addition to enhancing the chair’s mobility, the legs provide additional traction on unprepared and slippery surfaces. The controller uses foot force information to coordinate the actuators of the legs and wheels so that the tendency to slip is minimized. The hybrid system is more attractive than a walking chair because it relies on wheeled locomotion that is established to be reliable and safe. The legs are used as crutches and only when they are needed. Furthermore, because the legs are not used to support the entire weight of the chair, the motors, controllers and the legs can be made as compact as needed. When the legs are not required for support, they can be used as manipulators to push open doors, reach for objects and move obstacles out of the way. When they are not needed, they are tucked away below the arm rest to make them inconspicuous. However, unlike a legged system, the hybrid chair cannot locomote without wheels. The reduced complexity, lower cost, and improved reliability and safety is at the expense of some loss in mobility. An important design consideration is the aesthetics of the design and consumer acceptance. The disadvantage of employing a fundamentally different method for locomotion is that the user may feel conspicuous using such a chair. While this distractibility factor depends to

Does not generalize to other environments, does not work on all types of stairs, often a bulky addition to the wheelchair, slow to deploy, poor maintenance. Requires transfer. Suitable for only small obstacles, due to power limitations of the wheelchair.

a large extent on the environment and society, it is necessary to make any design more un-robot-like. TELETHESES AND HUMAN EXTENDERS: RESEARCH ISSUES The discussion on prosthetics revealed two essential features for a successful design. These are (a) a three-dimensional, one-to-one, map between the user’s input motion and the manipulator’s motion and (b) force reflection from the manipulator to the user. Bilateral control provides for extended physiological proprioception (EPP) and this allows for superior control and performance. In the discussion that follows, we look at a class of devices that can be considered as extensions of prosthetic limbs. Like prosthetic devices, they are intimately linked to the human user and enable EPP. Furthermore, they are passive and powered by the human user, although they may include electromechanical, power-assist mechanisms. However, unlike prosthetic devices they may possess more than two degrees of freedom and are more reminiscent of robot manipulators. We first look at feeding aids as examples of such devices and then describe research prototypes of more complex, general-purpose aids.

Figure 2. (a) CAD model and (b) photograph of the hybrid all-terrain wheelchair developed at the University of Pennsylvania.


709

Feeding Devices There is a high degree of motivation for people to learn to feed themselves. A recent survey of the U.S. population indicates that the population indicates that the population that would benefit from well-designed feeding aids may be as high as half a million. There are commercially available feeders that are useful for people who have controlled movements of the head and neck and can take food off of a feeding utensil that is brought close to the mouth. The Winsford Feeder (Winsford Products, Pennington, NJ) and the Beeson Feeder (Maddox Inc., Pequannock, NJ) are two such examples. Another example is the Handy I, a robotic arm that is programmed for feeding. Most feeders consist of an articulated, electrically powered arm with a spoon at its end, a plate on a rotating turntable and an auxiliary arm that may be used to push food on to the spoon. The user controls, through the use of switches, the movement of the different components. Although such feeding aids can be used effectively, there are several reasons why their use is not as widespread as one would expect (74). The control switches may initiate a movement of a certain component, for example, a rotation of the plate. The user may find it difficult to stop the motion. Visual feedback is required for successful feeding during most of the operation. It is acceptable if vision is required to locate the morsel of food and to target it. But requiring the user to focus on the morsel through the entire operation of scooping it up and bringing to the mouth can be very exhausting for most users. Some devices require an electrical outlet and this may be a nuisance factor. Finally, they are expensive and are difficult to adapt to individual needs. A completely different solution is exemplified by the Magpie (75) shown in Fig. 3. The ankle, knee, and thigh movements are coupled via a set of cables and pulleys to a four degree-of-freedom articulated arm. By moving his or her leg in a controlled manner, the user can feed effectively. Because it is physically and intimately coupled to the user and acts as an extension of the person, such a device is called a telethesis. A telethesis has the flavor of a prosthetic limb (except for the number of degrees of freedom) and therefore the user is always in intimate contact with the limb. This offers the user a form of proprioceptive feedback. Because of this intimate physical contact, the user will always know the position and orientation of the spoon and the articulated arm and will only use vision to locate the target morsel. Furthermore, such devices are simple, reliable, and inexpensive and may not require actuators. Clinical trials show a high degree of consumer acceptance (75). However, since the target population consists of users that have limited upper extremity movement but intact musculature in their legs, its usefulness is rather limited. The prototype in Fig. 4 is a telethesis that uses head and neck movements to control the movement of a spoon. The linkage has three degrees of freedom and in particular, is capable of three distinct output motions. It can be used to scoop up the food from the plate with any approach angle and bring the food to the user’s mouth as he or she pitches his or her head forward. The mechanism has three degrees of freedom driven by cables. The nominal yaw movement of the head, causes the linkage to rotate about a vertical axis and translate in a horizontal plane so that the spoon is always in the line of the sight of the user. The nominal pitch movement of

Figure 3. An articulated mechanism for feeding in a foot-controlled feeding device (Magpie) designed at the Nuffield Orthopaedic Center in Oxford, England (75).

the head drives a planar open chain (whose joints are coupled linearly) so that the spoon performs a planar motion that involves scooping up the food and bringing it to the mouth. The nominal roll movement, causes the spoon to pitch about a transverse axis. Such passive mechanical feeders can be less expensive and easier to operate than electrical feeders. The main concern is that the prototype in Fig. 4 has to be worn by a user and looks like a mechanical aid. While this may not be a concern in a dining hall or in a home, it may not be socially acceptable. In contrast, the prototype in Fig. 5 has fewer components and is not worn by the user. Thus the user may detach himself from the device for social interactions. However, the lack of the physical coupling at all times may also be a potential disadvantage because the EPP link is broken. The spoon assembly is supported by a gravity compensated mechanical arm. The user uses his or her mouth to manipulate the spoon directly and to rotate the plate. A mechanical clutch locks the spoon while the user rotates the spoon about a vertical axis to bring the scooped food to the mouth. Body-Powered Manipulators Another example of a telethesis is the Chameleon (78), a wheelchair-mounted, counterbalanced, electromechanical arm. The arm’s end point is controlled and/or powered by a

710


Figure 4. A feeding device designed at the University of Pennsylvania (76). (a) CAD model and (b) photograph of the prototyped feeder.

functional body part of the user, via Bowden cables and/or an electric motor. The system consists of three main components: 1. The ‘‘slave’’ arm unit 2. The ‘‘master’’ or interface unit 3. The transmission and control systems The user engages the master unit, in this case through biting, and moves his or her head to control the slave arm. A transmission and control system connect the master to the slave so that the slave follows the master in a natural and intuitive manner. Two of the degrees of freedom between the units are connected through pulleys and Bowden cable; the third coupled degree of freedom is through an electric motor and controller.

The slave arm is shown in Fig. 6. All ␪ values represent angular motion about a corresponding axis; all ␳ values represent linear motion along an axis. This unit is mounted to a wheelchair such that joint ␪sp is at shoulder height and the unit is to the right side of the user. The slave arm was mechanically designed with the same degree of freedom as a spherical coordinate system: ␪sp (pitch), ␪sy (yaw), and ␳s (radial). A spherical coordinate system was chosen because it allows any position in space to be obtained with variables (inputs) that kinematically match a person’s input (in this case, head) and arm movements. Pulleys are located at joints ␪sp and ␪sy which are connected, via cable, to pulleys located on the master. A motor is present at joint ␪s␳; this along with the two meshing gears fixed to the ends of the two main links result in radial motion, ␳s. The slave arm is counterbalanced so that the user does not feel its weight when static. Roller bearings are present in all joints. The two main links are constructed of fiber-

Figure 5. A feeding device designed by a team of students from Cooper Union, New Jersey Institute of Technology, Ohio State University and the University of Pennsylvania (77). (a) CAD model and (b) photograph of the prototyped feeder.


θ sρ θ sp

ρs

Attached to wheelchair

θ sy

Distal end Figure 6. Slave arm of the Chameleon.

glass hollow tubing, while most other components are machined from aluminum. One of the interface or master units is detailed in Fig. 7. The unit is fixed to the user’s wheelchair such that the ␪my joint is positioned above and approximately at the center of

θ my

711

the user’s head. This head/mouth interface unit has four degrees of freedom; the three (␪mp, ␪my, ␪m␳) that map to the slave unit and an additional passive joint (␪mr) at the mouthpiece so the master unit does not constrain the natural head movement of the user. The mouthpiece is constructed of Polyform威 (distributed by Smith & Nephew, Inc.); a thermosetting plastic allowing the user’s dentition to be molded at low temperature. Roller bearings are present at the yaw (␪mp) and pitch (␪my) joints. Pulleys are present at the ␪my and ␪mp joints, while a rotary potentiometer is placed at ␪m␳. This potentiometer measures the translation of the mouthpiece, ␳m, which ultimately controls ␳s. Although not shown for the purpose of clarity, Bowden cables run between the master and slave units. Two sets of Bowden cables connect the yaw and pitch pulleys of the master and slave system; that is, ␪mp is connected to ␪sp and ␪my is connected to ␪sy. This set-up causes proportional (in this case, equal) angles of rotation between the two unit’s pitch and roll joints. Bowden cables are required, because relative movement occurs between pulleys of the master and slave units. Bowden cable is comprised of a flexible outer housing, a flexible steel wire, and a polyethylene liner that lies in-between the wire and outer housing to reduce friction. The radial position of the slave, ␳s, is achieved through closed-loop position control. An electric motor rotates joint ␪s␳ such that the error between the input (master) ␪m␳ and the output (slave) ␪s␳ is minimized. External power (a motor) was used at this degree of freedom to allow the person’s small translational head input motion to control the larger radial slave motion while still maintaining adequate force capability at the distal end of the slave. The advantages of this system are its ability to provide EPP and force reflection and its simplicity relative to rehabilitation robots. Although its complexity is a little more than that of the feeders discussed earlier, it is more versatile and allows a person with no or very little arm function to interact with his or her surroundings. The target population that would benefit from such a device is very large because the basic ideas can be adapted to any special purpose task (feeding is an example) and to other input sites (only head control is discussed here). Power-Assist in Worn Assistive Devices

θ mp ρm

θ mr

θ mρ

Figure 7. Head/mouth master interface of the Chameleon.

In many cases, it is desirable to provide a power-assist mechanism that can augment human power, much in the spirit of power-assist controls in automobiles and aircrafts. The first examples date back to the first teleoperators (devices that allows an operator to perform a task at a distance, isolated from the environment that the task is performed in) developed by Goertz (79) for manipulating radioactive materials. The next significant development can be seen in Mosher’s work (80) in the 1960s. He developed the Handyman, a master–slave manipulator for handling radioactive equipment. This work led to the development of a master-slave exoskeleton system called the Hardiman that allowed the human user to amplify his or her strength. Even in these early prototypes, the need for proprioceptive feedback and the need to reduce the number of degrees of freedom and simplify the coordination task were clearly understood. However, because they were master-slave systems, the human user was not in direct

712


Motion capture

Geometry capture

Central Interface Mechanism Design Module

Optimal kinematic synthesis

PRO/Engineer Visualization and user interface

Type synthesis Jack

Kinematic analysis

CAD modeling

Dynamic analysis

Manufacturing interface and postprocessor

Designer Manufacturing system CNC machine Figure 8. Flowchart for design and rapid prototyping of one-of-a-kind rehabilitation products.

contact with the manipulator or the leg. There was an electronic link that did not allow proprioceptive feedback. The ideal power-assist mechanism acts as an amplifier while allowing the human to remain in direct contact with the manipulator, thus enabling extended physiological proprioception (81). The basic underlying idea is to use force sensors to infer the force applied by the human and the force required for the manipulative task and supply the difference using an electromechanical actuator. A variation of this idea can be used to supress tremor or spasticity. Since the force applied by the human is sensed, it can be appropriately filtered before being used as an input signal. One of the main disadvantages is that the human user must interact physically with an actively controlled system, and there are concerns about safety and consumer acceptance. The control system must be designed so that the human–machine system remains stable under all conditions. One way of approaching this is by requiring that the electromechanical aid remain passive under all conditions (82). Because the device must interact with different conditions and the user’s condition may change over time, this is a challenging research problem with the potential of a great payoff. DESIGN AND MANUFACTURE The design and manufacture of assistive devices presents a novel problem. Because each person presents a unique neurophysiological picture, there is considerable variation of performance and function and therefore, it is essential to design tools that are specific to that person. It is necessary to involve the customer in any design process, but this is especially true

for rehabilitation aids. Furthermore, there are biological changes that occur over time, and it is necessary to allow for adjustments and maintenance or to rapidly redesign and manufacture a new product. Traditional models for product development and manufacturing focus on low-cost, high-volume products. In contrast to this, the manufacture of rehabilitation aids requires the infrastructure and technology to design and produce a wide array of quality products each of which targets specific market needs. Even though agile manufacturing (83) makes it possible for a designer to move quickly from a preliminary design concept to a prototype, it does not specifically address the need to customize products to individuals. Regardless of the specific product class, the first important step in the production of a customized product is the quantitative assessment of the needs of the individual. This involves the acquisition of individual geometric, kinematic, dynamic, and physiological information, which is necessary for developing design specifications and for detailed design. Because the product volume for customized products is likely to be small, the manufacturing cost must be kept low. Thus, there is a need to automate the process of measuring the customer and designing the product from specifications derived from these measurements. In addition, there is always pressure to provide the product quickly and be able to respond to the consumers’ needs rapidly. The design process for rehabilitation products that are customized to a person will involve a number of steps (84), as shown in Fig. 8. Of these, there are three stages that are particularly important for such products: data acquisition, virtual prototyping, and rapid design and prototyping.


713

Figure 9. The central virtual user interface for visualization and optimization of the design.

Data Acquisition It is necessary to measure the capabilities and needs of the individual, his or her environment, and to describe the task in quantitative terms in order to generate the specifications for the design problem. For example, the custom design of a head-controlled telethesis for feeding requires the measurement of the geometry of the head, the kinematics of the head and neck, and the forces that the person can apply with his or her head. Similar measurements may also be required for the feeding task (e.g., the ranges of motion of the spoon or fork and the forces that are encountered during the task). For customized design, we require, in addition to geometric measurements (shape, size), information about the kinematics and dynamics of the individual. Virtual Prototyping Virtual prototyping is the process of design, analysis, simulation, and testing of a product within the computer and using the results to refine the concept and redesign the product before making a physical prototype. Over the last decade, highspeed computer graphics workstations have proven to be very

effective in allowing visualization of three-dimensional complex systems (85). With advances in robotics technology, the potential for developing haptic interfaces that allow the user to feel forces exerted by the virtual environment (in addition to seeing the environment) has been successfully demonstrated (82). As computers become faster and as more sophisticated actuators and sensors are developed, computer interfaces will enable the user to feel, touch, and see the virtual product in a virtual environment. For customized design and prototyping, it is essential to integrate virtual prototyping with data acquisition. With the measurement of the user, the task, and the environment, we can create accurate dynamic models (specific to the user, the task, and the environment) and investigate the virtual creation and installation of a customized virtual product on a vritual human user as an integral part of the engineering process. Consider the example of a feeding device. To evaluate candidate designs, it is useful to create a simulation of the user and the mechanical system as shown in Fig. 9. The mechanism that links the human head to the feeding device is not shown in the figure. The designer can experiment with differ-

714


ent kinematic coupling mechanisms and see how the movements of the user are translated into the movement of the end effector or the spoon. Three-dimensional graphics provide visual information about the design, whereas a real-time dynamics simulation package elicits information about the forces and the velocities that are required of the human head and neck to effectively accomplish feeding. By linking to an appropriate physiological database one can verify the feasibility of the required head and neck motions and also investigate possible sources of discomfort or trauma with the virtual prototype before clinical tests are performed. Being able to develop a virtual prototype of the product also allows the consumer to use and evaluate the virtual product in an appropriate virtual environment before the designer commits to the expense of creating the physical prototype. As shown in Fig. 9, consumer feedback (and evaluation by experts such as therapists) during the virtual prototyping phase and the redesign of the product in response to this feedback at a very early stage can ensure the success of the product and possibly avoid building multiple physical prototypes and incurring the resulting expenses. Rapid Design and Prototyping The design process can be divided into a concept development and system-level design phase and a detail design phase (86). By rapid design we mainly refer to speeding up of the detail design phase, the process of taking a preliminary design, converting it into a detailed design to quickly produce a prototype for evaluation and testing. It includes the specification of the geometry, materials, and the manufacturing process for each component. The key to speeding up the process is integration, in this case between virtual prototyping and rapid physical prototyping. This allows the designer to ‘‘kick the tires of the product’’ before committing to manufacture. This integrated approach to design requires a sophisticated computer interface that allows the designer to access various heterogeneous pieces of information. At the heart of our design package (87) is a graphic–user interface, which also acts as a server to support the interactive design and analysis processes. The key idea is to have a generic request procedure that enables any of the component design/analysis packages or modules to call another package to obtain relevant information. Thus information from any data acquisition, virtual prototyping, or simulation module can be displayed on the visualization package easily. Finally, since the modules operate on different machines/architectures, efficient communication protocols between separate processes (relying on Unix TCP/IP calls) are employed. Thus, this graphic server allows a modular approach to software development and enables the human designer to interact with each module at different levels. CONCLUSION We have presented a review of the technology underlying assistive devices for people with manipulative and locomotive disabilities focusing on prosthetic limbs, robotic arms, and wheelchairs. We have pointed out the important role that robotics can play in assistive devices. There is another class of assistive devices, called teletheses, that bear a strong resemblance to the multiple-degree-of-freedom robot arms and to

the body-powered prosthetic limbs. We discussed how a telethesis may be an optimal compromise that allows for extended physiological proprioception as well as strength enhancement. Finally, we discussed the design and manufacturing issues for such devices. The high degree of customization that is required and the one-of-a-kind flavor of these products suggest that a computer integrated, automated approach to design and prototyping is necessary for manufacturing. BIBLIOGRAPHY 1. C. Butler, Effect of powered mobility on self-initiated behaviors of very young children with locomotor disability, Devel. Med. Child Neurol., 28: 325–332, 1986. 2. A. L. Muhlenberg and M. A. LeBlanc, Body-powered upper-limb components. In D. J. Atkins and R. H. Meier (eds.), Comprehensive Management of Upper-Limb Amputee, New York: SpringerVerlag, 1988. 3. D. S. Childress, Closed-loop control in prosthetic systems: historical perspective, Ann. Biomed. Eng., 8: 293–303, 1980. 4. D. S. Childress, Historical aspects of powered limb prostheses, Clin. Prosth. Orthot., 9 (1): 2–13, 1985. 5. R. N. Scott et al., Sensory-feedback system compatible with myoelectric control, Med. Biol. Eng. Comput., 18 (1): 65–69, 1980. 6. A. H. Bottomly, Myo-electric control of powered prostheses, J. Bone Joint Surg., 47B (3): 411–415, 1965. 7. N. Hogan, A review of the methods of processing ems’s for use as a proportional control signal, Ann. Biomed. Eng., 4 (1): 1976. 8. D. C. Simpson and J. G. Smith, An externally powered controlled complete arm prosthesis, J. Med. Eng. Technol., 275–277, 1977. 9. M. Soede, Mental control load and acceptance of arm prosthesis, Automedica, 4: 183–191, 1982. 10. D. E. Whitney, Resolved motion rate control of manipulators and human prostheses. IEEE Trans. Man–Mach. Syst., MMS-10: 47– 53, 1969. 11. C. J. Abul-haj and N. Hogan, Functional assessment of control systems for cybernetic elbow prostheses, parts i–ii, IEEE Trans. Biomed. Eng., 37: 1025–1047, 1990. 12. G. F. Shannon, Some experience in fitting a myoelectrically controlled hand which has a sense of touch, J. Med. Eng. Technol., 2 (6): 312–314, 1978. 13. R. W. Witra, D. R. Taylor, and F. R. Finley, Pattern recognition arm prosthesis: a historical perspective—a final report. Bull. Prosth. Res., 10 (29): 8–36, 1978. 14. D. W. Lamb et al., The management of upper limb deficiencies in the thalidomide-type syndrome, J. Roy. Coll. Surg. Edinburgh, 10: 102–108, 1965. 15. C. A. McLaurin, Control of externally powered prosthetic and orthotic devices by musculoskeletal movement, in The Control of External Power in Upper Extremity Rehabilitation, Washington, DC: National Academy of Sciences, 1966. 16. E. G. Johnson and W. R. Corliss, Teleoperators and human augmentation, AEC-NASA Technology Survey, December 1967. 17. B. Hannaford, Stability and performance tradeoffs in bi-lateral telemanipulation, in Proc. 1989 IEEE Conf. Robot. Autom., 1989, pp. 1764–1767. 18. N. Hogan, Impedance control: an approach to manipulation, parts i–ii, J. Dynamic Syst. Meas. Control, 107: 1–16, 1985. 19. D. C. Simpson and D. W. Lamb, A system of powered prostheses for severe bilateral upper limb deficiency, J. Bone Joint Surg., 47B (3): 1965.

ASSISTIVE DEVICES FOR MOTOR DISABILITIES 20. R. E. Prior and C. M. Scott, Proportionally controlled linear power assist device for artificial arms, Bull. Prosth. Res., 10 (24): 43–50, 1975. 21. M. D. O’Riain and D. T. Gibbons, Position proprioception in a microcomputer-controlled prosthesis, Med. Biol. Eng. Comput., 25: 294–298, 1987. 22. C. W. Heckathorne, J. S. Strysik, and E. C. Grahn, Design of a modular extended physiological propioception controller for clinical applications in prosthesis control. Proc. RESNA 12th Annu. Conf., Washington, DC, 1989. 23. J. A. Doubler and D. S. Childress, An analysis of extended physiological proprioception as a prosthesis-control technique. J. Rehab. Res. Devel., 21 (1): 5–18, 1984. 24. J. A. Doubler and D. S. Childress, Design and evaluation of a prosthesis control system based on the concept of extended proprioception, J. Rehab. Res. Devel., 21 (1): 19–31, 1984. 25. J. R. Allen, A. Karchak, and V. L. Nickel, Orthotic manipulators, in Advances in External Control of Human Extremities, Belgrade, 1970. 26. V. Paeslack and H. Roesler, Design and control of a manipulator for tetraplegics, Mechanism Mach. Theory, 12: 413–423, 1977. 27. C. P. Mason and E. Peiser, A seven degree of freedom telemanipulator for tetraplegics. Conf. Int. sur les Telemanipulators pour Handicapes Physiques, 1979, pp. 309–318. 28. W. Seamone and G. Schmeisser, Early clinical evaluation of a robot arm/worktable system for spinal-cord-injured persons. J. Rehab. Res. Devel., 22 (1): 38–57, 1985. 29. G. Verburg et al., Manus: The evolution of an assistive technology, Technol. Disability, 5 (2): 217–228, 1996. 30. M. Topping, Handy I, a robotic aid to independence for severely disabled people, Technol. Disability, 5: 233–234, 1996. 31. C. Upton, The RAID workstation. Rehab. Robot. Newsl., A. I. duPont Institute, 6 (1): 1994. 32. H. F. M. Van der Loos, VA/Stanford rehabilitation robotics research and development program: lessons learned in the application of robotics technology to the field of rehabilitation, IEEE Trans. Rehab. Eng., 3: 46–55, 1995. 33. G. E. Birch et al., An assessment methodology and its application to a robotic vocational assistive device, Technol. Disability, 5 (2): 151–166, 1996. 34. S. J. Sheredos et al., Preliminary evaluation of the helping hand electro-mechanical arm, Technol. Disability, 5 (2): 229–232, 1996. 35. Teitelman, De-handicapping the handicapped, Forbes, 1984. 36. C. A. McLaurin and P. Axelson, Wheelchair standards: an overview, J. Rehab. Res. Devel. (Clin. Suppl.), 27 (2): 100–103, 1990. 37. T. K. K. Koo, A. F. T. Mak, and Y. L. Lee, Evaluation of an active seating system for pressure relief, Assistive Technol., 7 (2): 119– 128, 1995. 38. Anonymous, Tom Houston is a real stand-up guy, thanks to the versatile vertical wheelchair he devised, People Weekly, 32: 91– 92, 1989. 39. IMEX Riser Wheelchair. Product Literature, Imex Medical Inc., San Jose, CA. 40. Standup Wheelchairs. Product Literature, Levo Inc., Switzerland. 41. H. F. M. Van der Loos, S. J. Michalowski, and L. J. Leifer, Development of an omnidirectional mobile vocational assistant robot. In Proc. 3rd Int. Conf. Association Advanced Rehab. Technol., Montreal, June 1988. 42. R. Walli, DOE technology to develop TRANSROVR—omnidirectional wheelchair, DOE News Brief, October 10, 1996. 43. H. Hoyer, The OMNI wheelchair, Service Robot, 1 (1): 26–29, 1995.

715

44. M. West and H. Asada, A method for designing ball wheels for omni-directional vehicles, 1995 ASME Design Eng. Tech. Conf., DAC-29, 1995, pp. 1931–1938. 45. F. G. Pin and S. M. Killough, A new family of omni-directional and holonomic wheeled platforms for mobile robots, IEEE Trans. Robot. Autom., 10: 480–489, 1994. 46. J. D. Nisbet, I. R. Loudon, and J. P. Odor, The CALL Centre smart wheelchair. Proc. 1st Int. Workshop Robot. Applications Med. Health Care, Ottawa, 1988, pp. 9.1–9.10. 47. D. A. Bell et al., The NavChair: an assistive navigation system for wheelchairs based on mobile robot obstacle avoidance. Proc. 1994 IEEE Int. Conf. Robot. Autom., San Diego, CA, May, 1994, pp. 2012–2017. 48. D. Miller and M. Slack, Design and testing of a low-cost robotic wheelchair prototype, Autonomous Robots, 2 (1): 77–88, 1995. 49. O. Neveryd and Bolmsjo¨, WALKY, a mobile robot system for the disabled. Proc. 4th Int. Conf. Rehab. Robot., Wilmington, June, 1994. 50. M3S: A general-purpose multiple-master multiple-slave intelligent interface for the rehabilitation environment, Working Draft ISO1716-17, International Standards Organization, 1995. 51. J. M. Ford and S. J. Sheredos, Ultrasonic head controller for powered wheelchairs. J. Rehab. Res. Devel., 32 (3): 280–284, 1995. 52. D. M. Brienza and J. Angelo, A force feedback jobstick and control algorithm for wheelchair obstacle avoidance. Disability Rehab., 18 (3): 123–129, 1996. 53. P. F. Muir and C. P. Neuman, Kinematic modelling for feedback control of an omnidirectional wheeled mobile robot, in I. J. Cox and G. T. Wilfong (eds), Autonomous Robot Vehicles, New York: Springer Verlag, 1990, pp. 25–31. 54. D. A. Bell et al., An identification technique for adaptive shared control in human-machine systems. Proc. 15th Annu. Int. Conf. IEEE Eng. Med. Biol. Soc., San Diego, CA, October 1993, pp. 1299–1300. 55. J. Borenstein and Y. Koren, Tele-autonomous guidance for mobile robots. IEEE Trans. Syst., Man Cybern., 17: 535–539, 1991. 56. R. Borgolte et al., Intelligent control of a semi-autonomous omnidirectional wheelchair, in Proc. 3rd Int. Symp. Intelligent Robot. Syst. ’95, SIRS ’95, Pisa, Italy, July 1995, pp. 113–120. 57. T. Houston and R. Metzger, Combination wheelchair and walker apparatus. U.S. Patent 5,137,102, 1992. 58. M. W. Thring, Robots and Telechirs: Manipulators with Memory, Remote Manipulators, Machine Limbs for the Handicapped, New York: Halsted, 1983. 59. D. R. Voves, J. F. Prendergast, and T. J. Green, Stairway chairlift mechanism. U.S. Patent 4,913,264, 1990. 60. B. Most, Stair-climbing wheelchair. Popular Sci., 230: 108, 1987. 61. Phoenix, The Climbing universal wheelchair, Product Literature, Tunkers Industries, Rochester, MI. 62. S. Hirose et al., Terrain adaptive quadru-track vehicle HELIOSIII. 9th Annu. Conf. Robot. Soc. Jpn., 1991, pp. 305–306 in Japanese. 63. S. Hirose, T. Sensu, and S. Aoki, The TAQT carrier: a practical terrain-adaptive quadru-track carrier robot, in Proc. IEEE/RSJ Int. Conf. Intelligent Robots Syst., 1992, pp. 2068–2073. 64. M. H. Raibert, Legged Robots that Balance. MIT Press, Cambridge, MA, 1985. 65. S. Hirose, A study of design and control of a quadruped walking. Int. J. Robotics Res., 3 (2): 113–133, 1984. 66. K. J. Waldron et al., Configuration design of the adaptive suspension vehicle, Int. J. Robot. Res., 3 (2): 37–48, 1984. 67. V. Kumar and K. J. Waldron, Actively coordinated mobility systems, ASME J. Mechanisms, Transmissions Autom. Design, 111 (2): 223–231, 1989.

716

ASYNCHRONOUS CIRCUITS

68. S-M. Song and K. J. Waldron, Geometric design of a walking machine for optimal mobility, ASME J. Mechanisms, Transmissions Autom. Design, 109 (1), 1987. 69. C-D Zhang and S-M. Song, Gaits and geometry of a walking chair for the disabled, J. Terramechanics, 26 (314): 211–233, 1989. 70. D. R. Browning et al., Legged mobility, a wheelchair alternative. http://bucky.aa.uic.edu:80/DVL/drew/leggs.html, 1988 71. P. Wellman et al., Design of a wheelchair with legs for people with motor disabilities, IEEE Trans. Rehab. Eng., 3: 343–353, 1995. 72. V. Krovi and V. Kumar, Modeling and control of a hybrid mobility system, ASME J. Mech. Design, 1997, submitted. 73. V. Kumar, P. Wellman, and V. Krovi, Adaptive mobility system. U.S. Patent 5,513,716, 1996. 74. R. Mahoney and A. Phalangas, Consumer evaluation of powered feeding devices. Proc. RESNA 96 Annu. Conf., Salt Lake City, Utah, June 1996. 75. M. Evans, Magpie: Its development and evaluation. Technical Report, Oxford, England: Nuffield Orthopeadic Center, Headington, 1991. 76. V. Krovi et al., Design and virtual prototyping of rehabilitation aids, 1997 ASME Design Eng. Tech. Conf., DFM-107, Sacramento, CA, 1997. 77. G. Kinzel et al., The use of desktop teleconferencing to coordinate a cross-university project. Proc. UPCAEDM, 1996. 78. S. Stroud, W. Sample, and T. Rahman, A body powered rehabilitation robot, in Proc. RESNA ’96 Annu. Conf., Salt Lake City, Utah, June 1996, pp. 363–365. 79. C. Goertz, Manipulators used for handling radioactive materials. In E. M. Bennet (ed.), Human Factors in Technology, New York: McGraw-Hill, 1963, ch. 27. 80. R. S. Mosher, Handyman to hardiman. SAE paper no. 670088, 1967. 81. H. Kazerooni and J. Guo, Human extenders. ASME J. Dynamic Syst., Meas., Control, 115 (2): 281–290, 1993. 82. J. E. Colgate and J. M. Brown, Factors affecting the z-width of a haptic display, in Proc. 1994 IEEE Int. Conf. Robot. Autom., San Diego, CA, May 1994, pp. 3205–3210. 83. J. H. Sheridan, Agile manufacturing: stepping beyond lean production. Industry Week, 242 (8): 3–46, 1993. 84. V. Kumar et al., Rapid design and prototyping of customized rehabilitation aids. Commun. ACM, 39 (2): 55–61, 1996. 85. N. I. Badler, C. B. Phillips, and B. L. Webber, Simulating Humans: Computer Graphics, Animation, and Control, New York: Oxford University Press, 1993. 86. K. T. Ulrich and S. D. Eppinger, Product Design and Development, New York: McGraw-Hill, 1995. 87. V. Kumar, R. Bajcsy, and W. Harwin, Design of customized rehabilitation aids, in Proc. 7th Int. Symp. Robot. Res., Munich, Germany, October 1995.

VIJAY KUMAR University of Pennsylvania

TARIQ RAHMAN Alfred I. duPont Institute and the University of Delaware

VENKAT KROVI University of Pennsylvania

ASSOCIATIVE MEMORY. See CONTENT-ADDRESSABLE STORAGE.


}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP


Wiley Encyclopedia of Electrical and Electronics Engineering Hearing Aids Standard Article Jaouhar Mouïne1 and Mohamad Sawan2 1University of Sherbrooke, Sherbrooke, Quebec, Canada 2Ecole Polytechnique de Montreal, Montreal, Quebec, Canada Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6604 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (236K)

●

●

● ●




Abstract The sections in this article are Auditory Pathways and the Hearing Process Hearing Disorders Hearing Aids Recent Developments of Hearing Aids About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.


❍

❍ ❍


HEARING AIDS

651

External Ear As shown in Fig. 1, the external ear includes the pinna, the ear channel, and the eardrum (tympanic membrane). The basic task of this part is to collect and guide sound waves to the middle ear. The external ear has a frequency response with a 10 dB to 15 dB gain over frequencies ranging from 1.5 kHz to 7 kHz and presents two resonant frequencies at 2.5 kHz and 5 kHz. Middle Ear

HEARING AIDS About 10% of the population suffers from some degree of hearing loss (1). The largest group (about 45%) concerns people older than 65 years. The group of 25 to 45-year-old persons includes up to 42%. One can classify different degrees of deafness by considering that the lowest stage is that of people having difficulty in understanding speech in a group conversation or among an audience listening to a speaker. The next stage concerns individuals that have difficulty hearing direct conversation, while the third stage involves individuals having difficulty hearing over the telephone but can hear amplified speech. The most severe stage is associated with individuals that cannot hear speech under any circumstances. Assuming that this classification concerns people that have acquired the hearing defect after learning to speak language by ordinary means, today all of them can benefit of a hearing aid to help them overcome their problem. Other individuals that are born deaf or who have acquired severe deafness sufficiently early in life to prevent them from learning speech through the usual means require special education techniques with associated devices and means. Throughout history, different devices have been invented to help hearing-impaired people. First versions of hearing-aid devices used in the 18th century were often tapers or horns that guide the sound toward the vertex or the point supplied with a small opening placed on the ear. In the beginning of the 20th century, the first electronic hearing-aids were designed, using the diode tube and the triode, and were based on the telephone principle. Since the 1950s the continuous progress in electronic and mechanical engineering has given rise to different wearable and miniature devices. In the 1980s, great advances in integrated circuit (IC) technology allowed the design of much smaller devices and under-the-skin implantable devices. Nowadays, the rapid progress in surgery, in IC technology, in material study, and in different other related fields have given access to very sophisticated devices for different degrees of hearing loss. AUDITORY PATHWAYS AND THE HEARING PROCESS The auditory system is one of the most wonderful and complex systems of the body. Witness its sensibility to the random motion of air molecules in contact with the eardrum and its capability to support sounds as loud as the noise of a jet engine, or its capability to distinguish sounds in a noisy environment as in a discotheque. The system can be divided into four basic parts: (1) external ear, (2) middle ear, (3) inner ear and (4) nervous pathways that leads information to the brain for interpretation.

The middle ear is a small cavity separated from the external ear by the eardrum and from the internal ear by a bulkhead comprising two apertures covered by a flexible membrane: oval window and round window. The middle ear cavity is connected to the throat by the Eustachian tube—a passage that allows maintaining the pressure equilibrium on both sides of the eardrum. The basic function of the middle ear is to adapt the air impedance to that of the liquids of the inner ear. This adaptation is performed by a system of three small bones: malleus, incus, and stapes, transferring mechanical vibrations caused by the sound wave on the eardrum to the oval window. This ossicular chain gives a mechanical amplification ratio of about 27 dB associated to the lever system that it forms and the surface ratio between the eardrum and the oval window (2). Two small muscles are also connected to this mechanical system to prevent damages when in the presence of loud sounds. However, these muscles attenuate low-frequency moderated sounds, and then act as a high-pass filter. Inner Ear The inner ear consists of a complex channel system grooved in the temporal bone and filled by a liquid called perilymph. This bony labyrinth contains a membranous one, filled by another liquid called endolymph. This system includes three basic parts: (1) semicircular canals, (2) the vestibule and (3) the cochlea. The two first parts are responsible of the head position and the body balance, while the last one deals with hearing. The cochlea is shaped as a snail’s shell, coiled upon itself for two and a half turns. Two membranes, called Reissner’s membrane and basilar membrane, divide the cochlea into three parallel canals: (1) the cochlear duct (limited by both membranes), (2) the vestibular canal (separated from the latter by the Reissner’s membrane), and (3) the tympanic canal (separated from the cochlear duct by the basilar membrane). The duct is filled by the endolymph and the canals contain the perilymph and they communicate at the top of the cochlea. The tympanic canal ends at the round window while the vestibular one begins at the oval window. Hence, the mechanical movements transmitted by the ossicles of the middle ear to the oval window membrane are converted into hydraulic pressure waves traveling through the vestibular and the tympanic canals and around the cochlear duct. This will consequently reproduce the mechanical movements on the round window membrane. The combined fluid displacements induce undulations of the basilar membrane, which supports a structure known as the organ of Corti, the most important element in the entire hearing mechanism. This element, lying along one side of the basilar membrane, contains the cells that convert hydraulic pressure into electrical impulses to be sent to the brain (3). These cells are supplied with exciting hairs

J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering. Copyright # 1999 John Wiley & Sons, Inc.

652

HEARING AIDS

Middle ear Ear canal Small muscles

Semicircular canals Oval window Round window Vestibulocochlear nerve

Cochlea

Figure 1. Structure of the human ear. There are three parts: (1) external ear, composed of the pinna, the ear canal, and the eardrum; (2) middle ear, containing the ossicles and the Eustachian tube; and (3) the inner ear, grooved in the temporal bone and mainly composed of the cochlea, which is connected to the auditory nerve fibers.

overhung by a flexible membrane called the tectorial membrane. Shearing action between the latter and the basilar membrane causes hairs to bend, and their cells then generate electrochemical signals for transmission by the auditory nerve to the central nervous system. Figure 2 represents a cross-section of the cochlea showing inner-ear structures. Nervous Pathways When the sound waves have been gathered by the external ear, and converted to mechanical energy, then to hydraulic pressure waves, and finally into electrical impulses by the

Stapes Eustachian tube Incus Vestibule Malleus

External ear Pinna External ear

Inner ear

middle and the inner ear, they are transmitted to the brain following the final link in the chain of hearing: the nervous pathways. As illustrated in Fig. 3, these pathways are enormously complex and contain different relay stations— cochlear nucleus, superior olive, inferior colliculus, and medial geniculate. There are also some descending nervous fibers, passing these relay stations and running from the brain back to the various parts of the ear. It is apparently the way that the brain directs partial or complete elimination of sound signals having no immediate importance.

Thalamus

Medial geniculate

Vertibulocochlear nerve

Temporal lobe Inferior collicus Tympanic canal (perilymph)

Lateral lemniscus Midbrain

Vestibular canal (perilymph)

Medulla Reissner's membrane Tectorial membrane Hairs

Cochlear nuclei

Organ of corti Cochlear duct (endolymph Cochlea Ear canal

Basilar membrane Cells

Superior olive

Spiral ganglion

Ossicles

Cochlear duct Eardrum Oval window Round window

Figure 2. A cross-section of the cochlea. Shearing action between the tectorial membrane and the basilar membrane causes hairs to bend. As a result, an electrochemical signal is generated to travel along the auditory nerve toward the central nervous system.

Organ of corti

Figure 3. Nervous pathways of the auditory system contain different relay stations and involve ascending and descending nervous fibers.

HEARING AIDS

HEARING DISORDERS

The second category of hearing impairments is related to the sensorineural defects. These can be caused by:

People with good hearing can detect a sound with intensity as low as 15 dB and tones at frequencies in the range of 16 Hz to 20 kHz. Figure 4 shows the threshold of hearing. The level of ordinary speech ranges from about 60 dB to about 80 dB over a frequency range running from about 100 Hz to nearly 8 kHz, with most used frequencies falling between 400 Hz and 3.4 kHz. The total deafness exists when sound cannot be heard at less than an average of 85 dB in speech frequencies (4,5). At first thought, deafness may seem to be the simple inability to hear sounds of normal loudness. As a matter of fact, defective hearing also alters the quality of sound. In fact, speech sounds may be distorted, becoming muddy and unclear in a way that different words may be confused and indistinguishable or completely unintelligible unless their context makes them clear. All hearing disorders can be classified into two categories: (1) conduction hearing losses and (2) sensorineural hearing losses. The first one is associated to the conductive structures of the ear, and so has its origins in the external and middle ears. Since these parts are dealing specially with amplification of the sound, the defects consist in a reduction of the sensitivity to all sounds independently of their frequencies. On the other hand, the sensorineural hearing losses arise in the inner ear or in the brain, as a result of a malfunction of some cells of the organ of Corti or some fibers of the auditory nerve or in the auditory cortex of the brain. Defects in these parts may affect hearing over all or a portion of audible frequencies. Conduction hearing losses can have different causes as: 1. Blockage of the ear canal 2. Infection of the middle ear 3. Limitation in the middle ear bones movement due to some liquid accumulated or a muscle rigidity 4. Rupture of the eardrum that cannot heal 5. Bone growth freezing the stapes movements (otosclerosis) 6. Stiffening of incudo-stapedial joint (ankylosis) 7. Loss of incudo-stapedial joint (arrosion) 8. Loss of incudo-malleal joint 9. Loss, rupture, or fixation of ligaments

140

Intensity (dB)

120

Threshold of pain

100 80 60

Ordinary speech

40 20

653

Threshold of hearing

0 0.02 0.05 0.1 0.2 0.5 1 2 Frequency (kHz)

5

10 20

Figure 4. The threshold of hearing and pain. These thresholds vary with the frequency of the sound. The black region is the ordinary speech domain.

1. 2. 3. 4. 5. 6.

Illness, such as scarlet fever or meningitis Sudden exposure to very loud noise, blast, or explosion Long-term exposure to a noisy environment Heavy drug use Presbycusis Congenital transmission

HEARING AIDS Depending on the hearing loss class, different means can be used to remedy the situation. The remedy can consist of a simple cleaning, a surgically replacement of some of the middle ear bones by synthetic pieces, or having recourse to a hearing aid. The technological advances that have taken place since the 1980s have provided numerous kinds of hearing aids. These device principles used depend on the degree of hearing loss. All of these corrective options can be classified as follows: 1. 2. 3. 4.

Sound amplifiers Middle ear implants and bone conduction devices Vibrotactile and electrotactile devices Cochlear prostheses

Sound Amplifiers The well-known sound amplifier placed in the external ear can be used only if the structures of the middle ear and the inner ear are not damaged. This kind of hearing aid is mainly composed of a microphone, an amplifier, and an output transducer (speaker). Its basic function consists of frequency shaping and amplification, in order to compensate for the deficiency in the individual hearing. Most of these devices are designed using traditional analog amplifiers. Today’s advanced technologies have allowed reduction in size and power-consumption needs, to the extent that they are available in different molds, which fit behind the ear, in the ear at the pinna, or inserted in the auditory canal. This cosmetic consideration, making them more and more invisible, is the feature that increases their popularity in the deaf community. As a result, manufacturer’s efforts have been invested in reducing device size rather than developing advanced signal processing, which would require a larger package. Thus these devices show many performance limitations that cannot be overcome without having recourse to modern signal processing techniques and sophisticated algorithms involving DSP. The limitations in typical sound amplifier hearing aids can be summarized in restricted dynamic range, distortion, restricted bandwidth, and especially difficulty in understanding sounds in noisy environments. In general, the signal-to-noise ratio (SNR) needed by a hearing-impaired person to give speech intelligibility in noise comparable to that for speech in quiet is greater than that required by a normal hearing person (1). The dynamic range in such hearing aid is bounded by noise at low sound levels and by amplifier saturation at high sound levels. A typical dynamic range is about 55 dB, which is about

654

HEARING AIDS

half the dynamic range of a normal ear. This can result in distortion for many sounds and can even affect monitoring of the user’s voice. On the other hand, the high-frequency response of a typical sound amplifier hearing aid tends to fall off rapidly above about 5 kHz. This high-frequency value is insufficient for optimal speech intelligibility or music appreciation. However, increasing this frequency may give rise to distortion and can lead to feedback problems in the hearing aid. The latter is a major factor that limits the maximum gain and degrades the device frequency response. The problems related to the acoustic feedback depend on how the hearing aid is fit in the external ear and are most severe at high frequencies since this is where we find the highest gain (6). Some recent digital devices are using a feedback cancellation technique to allow increasing the gain of the device (7,8). This technique consists of estimating the feedback signal and to subtract it from the microphone input. There are other techniques used to improve this kind of hearing aid. The so-called compression amplification is used to prevent amplifier saturation or to match the dynamic range of the amplified sound to that of the impaired ear. Currently used devices have up to three frequency bands on which one can apply this technique. At the same time, special attention is paid to minimize the attack time constants and the release times of electronic amplifiers to prevent their saturation (9). Many other techniques are also used to improve speech intelligibility by improving the SNR of different devices. However, results are still far from the ultimate goal, and improvements of the SNR didn’t show significant improvements in speech intelligibility (10).

movements of the magnetic alloy piece. This signal is conveyed to the implanted part by using a transformer with its secondary coil installed under the skin behind the ear in the mastoid region and connected to the coil responsible for magnetic alloy piece oscillations. These implantable pieces are totally biocompatible and do not necessitate any maintenance. They may be a wedge that will be fixed to the malleus, an incudo-stapedial ring, or a full incus replacement (15,16). The other category of this type of devices is based on the skull bone conduction. The basic principle of these hearing aids can be summarized in the conversion of sounds into vibrations communicated directly to the skull of the device user. This is achieved by using a bone screw that is osseointegrated into the mastoid region or in some cases placed in the jaw. In general, a delay of about four months before using the device allows the osseointegration process to have sufficient stress transfer from the implant to the bone without progressive movements or abrasion (17,18). As in the other systems, the device includes an external unit composed of a microphone, amplifiers, filters, power supply, and transducer monitoring the bone screw vibrations. In general, these vibrations are generated by a magnetic field varying according to the input signal. The transducer can be coupled to the bone screw directly or transdermally. In the case of a transcutaneous link a transformer with implanted secondary coil should be used to convey the magnetic field that will generate mechanical vibrations. However, for direct connection only one coil is sufficient to vibrate a piston attached to the bone screw. This kind of hearing aid has shown performances up to 20 dB better than the best sound amplifier devices (19).

Middle-Ear Implants and Bone Conduction Devices This kind of hearing aid can be used if the inner-ear structures are still viable. In this situation an implantable device is appropriate in the case where the hearing disorder cannot be remedied by a sound amplifier or when the latter shows very weak performances or medical contraindication. Among the problems that can be encountered with a sound amplifier and overcome by these devices are low fidelity, feedback, poor frequency response, and allergic reactions to the sound amplifier hearing aid packaging to be worn, such as ear molds (11,12). The basic principle of the middle-ear implants consists of a controlled amplification of the ossicular chain movements. There exist two methods to achieve this. The first method uses a vibrator connected directly to the malleus or the stapes and the second one consists of controlling the ossicle movements by means of magnetic alloy pieces grafted to the ossicles. In either case electrical energy is transformed into mechanical energy. The basic system of the first type of implant consists of a microphone placed in the external ear canal, an amplifier, a battery that can be transdermally charged, and a piezoelectric transducer (vibrator). The latter converts the voltage excitation into mechanical displacements (13,14). The second way to control ossicle movements consists of using a coil implanted in the mastoid region near the middleear cavity to create an electromagnetic field that vibrates the ossicle supplied by the magnetic alloy piece. In this device an external unit including a microphone, an amplifier, and a battery, generates the signal that will control the mechanical

Vibrotactile and Electrotactile Devices The basic principle of this kind of device is based on the fact that the skin is a spatially extended sense organ as is the eye and have some temporal processing capabilities similar to the ear. In fact, most of the qualities of the sensation produced when the ear is stimulated can be associated to similar counterparts in the skin sensation. Although these hearing aids are classified as assistive listening instruments, their performances demonstrated surprising results for sound and speech recognition when used alone (20,21). Different schemes have been designed to be used as tactile hearing aids. They range from single-channel aids, providing minimal information on the sound and using simple processing of the speech signal, to multichannel ones, which incorporate frequency vocoders and use complex processing and sophisticated speech processing algorithms to bring out and provide as many sound features as possible (22). Regardless of the complexity of the device and the strategy of the signal processing, the aim of these hearing aids is to translate sounds into sensations to be perceived on any part of the deaf individual’s body. The tactile hearing aids can be divided in two classes. The first one includes vibrotactile devices, in which the acoustic signal is presented as a vibration to the skin by using a mechanical transducer. The second class involves electrotactile devices, in which the acoustic signal is converted to electrical current pulses stimulating the skin. One of the most important factors that must be taken into account is that the skin does not respond well to stimulation frequencies much above

HEARING AIDS

1. Onset and duration of the fundamental frequency of voicing and formant information 2. Rhythm of speech 3. Duration of a signal 4. Limited changes in loudness 5. Discrimination of common environmental sounds 6. Feedback regarding production parameters of the user’s own speech 7. Complement to the visual sense (as in lip-reading), by coupling this skill with information regarding frication, nasality, and voicing Despite all the benefits that can be accomplished by this kind of hearing aid, there has been no widespread use of them except for individuals who are blind besides being deaf. This is probably due to the fact that they are not using any part of the auditory system. This may be unacceptable for deaf individuals, especially when they are convinced that they can correct their hearing loss instead of looking for another means, different from that of a normal person, to hear. On the other hand, wearable tactile transducers may be reason enough to discourage deaf individuals from using these devices. Cochlear Prostheses As already mentioned, the final link in the chain of hearing is the nervous pathways. This means that, as in any other sense organ, the role of the ear consists in transforming the acoustic signal to electrical pulses traveling along nerve fibers toward the brain to be interpreted. The idea behind the use of cochlear prostheses is to deliver electrical pulses directly to the nerve fibers, according to the acoustic signal to be perceived. Thus these devices are intended for people suffering from sensorineural hearing loss and recognized as being totally or profoundly deaf. Since this kind of device is designed to replace a major part of the hearing system, it’s the most complex and the most

From a microphone

External part Coding and/or modulation

Sound analysis Skin Microstimulator

Demodulation and/or decoding

Communication link

600 Hz. Thus a special signal processing must be accomplished, to include high-frequency information of the sound inside this frequency band. Since the perception of sounds is no longer associated to the hearing system, this frequency transposition together with noise suppression techniques that can eliminate most of ambient background signal, give to these devices a great advantage over the bone conduction devices. According to these considerations, a simple system of a tactile hearing aid involves a microphone, power supply, amplifiers, filters, frequency transposition circuits, noise suppression circuits, and a transducer, which can be a mechanical vibrator or an electrode array delivering a current source electrical stimulus. Of course, a more complex system can include a DSP to perform more sophisticated signal processing strategies to extract different sound features. In either case the transducer is worn on the wrist and in some cases it can be located in other sites on the body as on the sternum or on the fingers. To sum up, a tactile hearing aid may be able to provide deaf individuals with a lot of information on the sound as:

655

Electrodes Implanted part Figure 5. The basic constituents of a cochlear prosthesis. The sound is collected by a microphone, then processed to extract its characteristics. The appropriate actions to be taken by the implanted part will then be dispatched via the communication link.

expensive of all hearing aids. As Fig. 5 depicts, the basic constituents of the system can be summarized in a sound analyzer, a stimulus generator (generally surgically implanted under the skin), a communication link between these two parts, and an electrode array that delivers electrical pulses to auditory nerve fibers. Today, there exist different systems that are used worldwide and that are subject to continuous improvements. Besides using different processing algorithms, these systems differ, especially by their implanted part concept. The latter can be monoelectrode or multielectrode supplied by extracochlear, intracochlear, or modiolar electrodes, using a percutaneous plug or a transcutaneous link and delivering monopolar or bipolar stimulation. All this diversity arises from the fact that the stimulation techniques that should give maximum speech recognition are still ignored. Hence, each device is based on its designers’ understanding of the human brain functions, and the way they try to emulate the external, middle, and inner ear functions. Regardless of the type of the device, results obtained by the same one with different individuals show considerable dissimilarities and complicate the situation by making each individual a unique case. Cochlear prosthesis can be divided in two big categories: (1) monoelectrode/single-channel and (2) multielectrode/ multichannel. The electrodes in both categories can be placed in different stimulation sites. When they are placed on the round window on the promontory or distributed over the cochlea, they are called extracochlear. In cases where they are inserted along the tympanic canal via the round window, they are called intracochlear. The last case corresponds to those impaled into the auditory nerve and called modiolar. Even though they can be justified by their contact with nerve fibers, modiolar electrodes are not suggested because of safety considerations since they damage the nerve and they risk affecting the facial nerve, which is very close to the auditory one. On the other hand, cochlear implants of both categories can either generate monopolar or bipolar stimulus. The first one is characterized by a common ground for all electrodes, that is, placed relatively far from the stimulation site. This mode can be used in case of small numbers of residual nerve fibers or hair cells since it spreads the stimulus over a large region. The bipolar mode consists of stimulating between two sites

656

HEARING AIDS

(positive/current source and negative/current sink) close to each other, which permits localization of the charge injection in a restricted region. With regard to the communication link between the sound analyzer and the implant, there exist basically two types. The transcutaneous one consists of an inductive link whose secondary coil is placed under the skin with the implant. The second one consists of a percutaneous plug and is less popular, since it exposes the individual to infection risks. However its transfer efficiency is far better than that of the first one. In both cases the communication link is intended to dispatch data and power to the internal part. The category involving single-channel cochlear prostheses owes its success to the simplicity of its concept. Figure 6 depicts the basic block diagram of a design example of such devices, which achieved great success (23). This hearing aid uses a single electrode inserted in the tympanic canal via the round window and delivering an analog output signal that consists of a sine wave whose amplitude is overmodulated by the band-limited acoustic signal. It is a well-established fact that in this way, there is no means to achieve maximum speech discrimination nor of hoping to get further performance improvements. However, the quasi-absence of any signal processing and the very simple concept of the design allow a considerable reduction of the size of the device to the extent that it can be worn behind the ear as a sound amplifier hearing aid. On the other hand, their simplicity limits their costs especially with regard to the function they are intended to achieve. These two considerations make them very popular and increase their success. The second category that includes multichannel cochlear implants offers much better performances and a lot of hope to reach optimum speech discrimination (24). This is due to their selectivity of the stimulated nerve fibers and the numerous possibilities of signal processing strategies that can be used for them. The most popular devices of this type are those using a powerful speech analyzer including a DSP, a transcutaneous communication link, and an intracochlear electrode array. However there is no consensus on the ideal number of electrodes needed. Because of their complexity due to the sophisticated operations that they are asked to perform and the different specialists (engineers, surgeons, psychologists, audiologists, speech-language pathologists, rehabilitation specialists, educators) involved to bring them into operation, these devices are the most expensive and necessitate a long

Sinusoidal oscillator

From a microphone Microphone preamplifier

BPF

Single electrode

Amplifier

Inductive link

Figure 6. The block diagram of a single-channel cochlear prosthesis. The stimulus is an analog signal that consists of an amplitude overmodulated sine wave.

rehabilitation time. Hence, their recipients should be carefully selected among the totally and profoundly deaf community. The selection criteria used by the majority of teams offering them are: 1. Being 18 years or older 2. Having acquired the hearing defect after learning to speak language by ordinary means 3. Being profoundly or totally deaf at both sides 4. Conventional hearing aids are not of any benefit 5. Being motivated to undertake a long rehabilitation program In general, in multichannel cochlear prosthesis the microphone is located at the ear level to detect acoustic signals. Then the latter is communicated to the sound analyzer (external part), which processes it and dispatches appropriate control data to the implant (internal part) via the communication link. The implant consists of an electrical stimulus generator and is usually placed behind the ear, under the skin in the mastoid region. The generated stimulus is then presented to the electrodes, which are inserted into the cochlea via the round window. Since the nerve fibers ending in the organ of Corti, which is lying on the basilar membrane, are selectively excited depending on the acoustic signal perceived, the electrodes are distributed along the tympanic canal, close to this membrane. They are addressed according to the sound characteristics tending to emulate the effect of the combined liquid movements and the response of the hair cells. It has been established that the stimulus should be a current waveform rather than a voltage one (25). This can be explained by safety considerations since the charge injected depends directly of the current level and thus can be better controlled. On the other hand, the current waveform used should be biphasic (completely balanced) to prevent any damage due to direct current accumulation and irreversible chemical reactions that occurs at the electrode–tissue interface. The most used current waveform consists of a rectangular pulse, allowing one to control the charge quantity by setting its amplitude and/or its width. To guarantee charge balancing, there’s often a series capacitor connected to the current source output. The ultimate goal of all cochlear implants is to find the ideal stimulation strategy that includes all of the speech signal features to provide all the information needed to the brain. The acoustic features of the sound waves have time and frequency specifications. The problem is that no one knows how these features should be presented to the inner ear, or what proportion of importance should be given to each one. Most of available speech processing algorithms are based on the three most important features: (1) peaks in vocal-tract transfer function (called formant positions), (2) vocal tract excitation rate (called pitch or fundamental frequency) and (3) the energy of the signal. Two basic classes of sound analyzer are used for these devices. The first class adopts an analog approach to process and to extract the signal characteristics. An example of such processors is shown in Fig. 7. This one, called a compressed analog processor, uses automatic gain control circuitry to achieve the dynamic range compression of the sound. The bandpass filters extract the fundamental and the higher for-

HEARING AIDS

Amplifier

BPF

CH1

From a microphone

657

Envelope detector

Spectral estimation

F0

BPF

Spectral estimation

F1

Spectral estimation

A1

Spectral estimation

F2

Spectral estimation

A2

From a microphone Automatic gain control

BPF

Amplifier

BPF

Amplifier

CH3

BPF

Amplifier

CH4

CH2

Automatic gain control

BPF

Figure 7. The block diagram of a compressed analog sound processor. The stimulus is delivered simultaneously over all channels.

mant information of the speech. The resulting signals are amplified and dispatched simultaneously to the cochlear electrodes (24,26). The problem that may be encountered when the stimuli are injected simultaneously is the interaction between the different channels inside the cochlea. Other processors propose an enhancement of this system to overcome this problem by using interleaved nonsimultaneous stimuli (26,27). These systems, called continuous interleaved sampling processors, have been designed with more than four channels to allow using more electrodes. Figure 8 shows such system. The half-wave rectifiers and the low-pass filters are used to extract the envelop of different signals collected at the output of different bandpass filters associated with each output channel. These envelop signals are then applied to a nonlinear mapping function to compress the dynamic range of the speech signal. Finally, each channel information modulates the amplitude of a biphasic current stream, including tempo-

Figure 9. The block diagram of a feature-based sound processor. Only the fundamental frequency, the first formant and the second formant characteristics, are estimated.

ral offset to modulate the current pulse position of channel relative to each other. The second class of sound processors is based on a digital approach. The system estimate some spectral features of the speech signal according to which the stimulus will be generated by the implanted part. Figure 9 shows an example in which the processor estimates the fundamental voice frequency (F0), the first formant frequency and amplitude (F1, A1) and the second formant frequency and amplitude (F2, A2) (28,29). F0 will be used to set stimulation frequency, F1 and F2 to address the affected electrodes and A1 and A2 to set the energy level. This concept has demonstrated some limitation,

BPF

Half-wave rectifier

LPF

Mapping function

Driver

CH1

BPF

Half-wave rectifier

LPF

Mapping function

Driver

CH2

. . . .

. . . .

. . . .

. . . .

. . . .

BPF

Half-wave rectifier

LPF

Mapping function

Driver

From a microphone Microphone preamplifier

. . . Pulse generator Figure 8. The block diagram of a continuous interleaved sampling sound processor. The current pulses are amplitude-modulated, including a temporal offset to prevent the interaction between channels.

CHn

658

HEARING AIDS

Microphone preamplifier

Sound compressor

BPF

Full-wave rectifier

LPF

BPF

Full-wave rectifier

LPF

. . . .

. . . .

. . . .

BPF

Full-wave rectifier

LPF

ADC

Microprocessor

From a microphone

Data and conrtol signals of the implant

Figure 10. The block diagram of a spectral maxima sound processor. An advanced device involving a modern DSP.

since it is based on only a few number of speech features. Other alternatives have been proposed to improve this kind of processor by subdividing the sound frequency band into more subbands over which the sound spectrum is evaluated and then the most significant results over all the sound frequency band are used to generate stimuli. This kind of sound analyzer is called spectral maxima sound processor and is depicted in Fig. 10 (30). In this system, the dynamic range of the sound is compressed at the input, then the signal is passed through an analog signal processing part, consisting of filters and rectifiers, to be converted to digital data and to be processed by a powerful microprocessor. It should be mentioned that, despite the complexity of these devices and the recent technological advances, there remains a lot of work to do to find the optimum stimulation strategies and to minimize their sizes. This will enable professionals to offer them to more deaf people including those born deaf and especially deaf children. RECENT DEVELOPMENTS OF HEARING AIDS The technological advances achieved since the invention of the transistor are progressing rapidly. The high-tech explosion of this era concerning all of the information technology systems is more and more present in daily life. As a result, a digital system revolution is extended by miniaturization, increase of circuit density and speed, and then by more sophisticated systems performing very complex computing and operations. This is directly reflected in the new and future hearing aids. The most affected systems are the sound amplifier devices and the cochlear prostheses. The former tend to extend their use by increasing their gains with better performance and the others should be accessible to more and other deaf people such as children, prelingually deafened, and even those who can obtain only marginal benefits from other amplification systems. Advanced Sound Amplifiers Hearing Aids Conventional sound amplifier devices that are using analog technologies cannot cope with intricacies of hearing and re-

spond only partially to the needs of deaf people. The new digital audio techniques made available through modern integrated circuits can match more closely the individual needs and situation. One of the major capabilities of these new designs is programmability, giving access to up to 100 softwarecontrolled parameters with regard to the maximum of 10 parameters available on their analog predecessors. Most designers working on this kind of hearing aid concentrate on improvement of the feedback cancellation, compression amplification, and noise suppression, in order to increase their gains and to maximize speech quality. Besides working on the same old concept to improve it with digital approaches, there are other designers exploring other concepts such as by using microphone arrays. The adaptive-feedback-cancellation technique, which consists of estimating the feedback signal and subtracting it from the microphone input, has shown 8 dB to 10 dB increase of the gain with digital designs (31,32). Furthermore, the latter allowed the implementation of multiband compression algorithms using different parameters to be chosen for optimum compression results. One objective of such systems is to place as much of the speech signal as possible within the residual hearing regions (33,34). A second objective consists in matching the loudness in the impaired ear to that in the normal one. All of these techniques have shown some improvements in speech quality (35,36). However, the major problem of intelligibility of speech in noise is still unsolved. Other means are used to treat this issue. In fact, instead of trying to reduce or to remove the noise, new techniques suggest enhancing the speech signal. This can be achieved by increasing the amplitude of spectral peaks while reducing the amplitude of the remaining frequency regions. This technique has not yet proven a significant enhancement of speech intelligibility, but many other techniques can be adopted, thanks to the flexibility and programmability of digital designs. To overcome the problem of improving the quality of speech in noisy environments, other techniques based on focusing on the desired signal at the microphone input lead to use directional ones. This idea uses the fact that in many situations the desired acoustic signal comes from a single welldefined source while the noise sources are located throughout

HEARING AIDS

the area. Thus a directional microphone or a microphone array, built into an eyeglass frame, for example, is used to maintain high gain in the direction of the desired signal and to reduce gain for other direction sources (37,38). By using different possible signal processing strategies, these techniques showed up to 15 dB improvement in SNR. On the other hand, when these techniques are used with adaptive arrays, they can give greater improvements in SNR than when using fixed weight arrays (39). Despite all of the research work undertaken to improve the quality of these devices, it still seems that because of the complexities of the auditory system and the nature of auditory impairments, advanced signal processing for hearing aids is a very difficult engineering problem. However, if deaf people can benefit even slightly from a new signal processing strategy, then this work is worthwhile. Present and Future Directions for Cochlear Prostheses One can say that cochlear prostheses have reached a robust childhood stage, since they have been shown to result in successful speech perception in profoundly deaf people. All of the specialists in auditory anatomy and physiology, otolaryngology, audiology, rehabilitation, education, speech-language pathology, bioengineering, psychology, and other related disciplines, as well as deaf people, are working together to extend their use and to make them accessible to a wider population of deaf people by establishing new selection criteria based on their success. The new digital processing techniques, together with the advanced integrated circuit technology, have allowed the implant and the sound processor sizes to be reduced, while at the same time increasing the power of the computing and processing strategies. At the beginning, cochlear prostheses were intended for postlingually deafened adults. However, experiments of recent years have demonstrated that even prelingually and perilingually deafened people can benefit from these devices. Furthermore, experiments have involved a relatively large number of children with prelingual, perilingual, and postlingual deafness and have demonstrated that young people should be considered among cochlear prostheses candidates. The minimum age of the cochlear prostheses recipients was 24 months, but results showed that the most important factor that can affect the performance of such devices is the early detection of the hearing defect. This is justified by better results obtained with people having shorter duration of deafness. On the other hand, it seems that younger age implantation may limit the negative consequences of auditory deprivation. These results in children have been mainly reported for single-channel devices or feature-based devices only. Knowing the limitations of this type of devices, a lot of hope remains in using other more advanced devices, of course, by making them suitable for children. Future developments of these devices concern all disciplines that are involved in their design stage, preimplantation stage, and postimplantation stage (40). Engineers should continue designing more flexible and powerful devices with respect to their safety considerations and by developing better signal processing and stimulation strategies to overcome the problem of noisy environments that still significantly detract from speech-perception abilities. Furthermore, since magnetic resonance imaging (MRI) is increasingly the diagnostic tool

659

of choice for a variety of medical conditions, providing better resolution of soft-tissue structures, the material of cochlear implants should be chosen to be compatible and then not to contain any magnetic or ferrous metals. Some other design consideration, such as providing self-test capability in the device, should overcome the problems related to failure detection, especially in young children. This complementary circuitry does not need to be very complex; a simple detection of electrode failure such as open and short circuits may be of considerable help. Specialists working with deaf people and involved in the preimplantation stage have much work to do to improve patient selection and future results. In fact, until now there is no residual hearing that is typically defined as profound hearing loss; moreover, the degree of preimplantation residual hearing does not predict postimplantation performances. Thus these specialists have to establish the critical distinction between the importance of residual pure tone sensitivity compared with that of overall residual auditory capacities. This should lead to well-defined audiometric criteria for candidacy, to better reflect functional auditory capacity. The postimplantation working specialists are concerned with reeducation programs and performance evaluation. The biggest challenge for those scientists is to find a tailored rehabilitation program for every individual and to develop protocols to reflect therapies effective for various types of individuals receiving implants. This is due to the fact that therapeutic intervention may differ significantly in time and content for prelingually, perilingually, and postlingually deafened recipients. Studies of the relationship between the development of speech perception and speech production in cochlear prosthesis users must be continued. On the other hand, special programs should be established for children, depending on their ages and including educational programs, taking into account the suitable auditory and speech instruction using the auditory information offered by the implant. Language acquisition should be an outcome measure in young children. The combined speech perception, language production, and language comprehension rehabilitation is particularly challenging and should give better results after the childhood stage. More comparative studies on language development in children with normal hearing, children with hearing impairment using conventional hearing aids, and deaf children with cochlear prostheses should be conducted. Other electrophysiological studies are also trying to solve the enigmatic side of the auditory system. Some of recent animal studies mention that electrical stimulation increases ganglion cell survival and also modifies the functional organization of the central auditory system. This issue should be developed to determine its implications in cochlear prostheses use. All of these efforts should bring further and more precise explanations on the wide variation in performance across individual cochlear prosthesis users. At the same time, every specialist in a well-defined feature of the cochlear prosthesis system should continue to improve it. Taking advantage of new low-power and reduced-size components, some of these efforts may offer more compact sound processors to improve at least the size of this part of the system by making it better hidden and (Why not?) behind the ear or in-the-ear wearable.

660

HEARING AIDS

BIBLIOGRAPHY 1. R. Plomb, Auditory handicap of hearing impairment and the limited benefit of hearing aids, J. Acoust. Soc. Am., 63: 533–549, 1978. 2. L. H. Roberts, Sound analysis by the ear, J. Med. Eng. Tech, 4 (4): 171–178, 1980. 3. A. Morgan, Donneés actuelles sur la Physiologie et la Pathologie de l’Oreille interne, Paris: Arnette, 1990. 4. M. C. Killion, Principles of high-fidelity hearing-aid amplification, in R. E. Sandlin (ed.), Handbook of hearing aid amplification, vol. I: Theoretical and Technical Considerations, Boston: Little Brown, 1988, pp. 45–79. 5. C. V. Pavlovic, Derivation of primary parameters and procedures for use in speech intelligibility predictions, J. Acoust. Soc. Am., 82: 413–422, 1987. 6. J. M. Kates, A computer simulation of hearing aid response and the effects of ear canal size, J. Acoust. Soc. Am., 83: 1952–1963, 1988. 7. D. K. Bustamante, T. L. Worrell, and M. J. Williamson, Measurement of adaptive suppression of acoustic feedback in hearing aids, Proc. Int. Conf. Acoust. Speech and Sig., 1989, pp. 2017– 2020. 8. O. Dyrlund and N. Bisgaard, Acoustic feedback margin improvements in hearing instruments using a prototype dfs (digital feedback suppression) system, Scand. Audiol., 20: 49–53, 1991. 9. E. Villchur, Signal processing to improve speech intelligibility in perceptive deafness, J. Acoust. Soc. Am., 53: 1646–1657, 1973. 10. P. Vary, On the enhancement of noisy speech, in H. W. Schu¨ssler (ed.), Signal Processing II: Theories and Applications, New York: Elsevier Sciences Pubs., 1983, pp. 327–330. 11. R. Goode, An implantable hearing aid, Trans. Amer. Acad. Ophthamol. Otol., 84: 28, 1970. 12. J. Vernon et al., Evaluation of an implantable type hearing aid by means of cochlear potentials, Volta Rev., 1: 20, 1972. 13. N. Yanagihara et al., Perception of sound through direct oscillation of the staples using a piezoelectric ceramic bimorph, Ann. Otol. Rhinol. Laryncol., 92: 223–227, 1983. 14. N. Yanagihara et al., Development of an implantable hearing aid using piezoelectric vibrator of bimorph design: State of the art, Otol. Head Neck Surg., 92 (6): 706–712, 1984. 15. K. J. Dormer et al., An implantable hearing device: Osseointegration of titanium-magnetic temporal bone stimulator, Am. J. Otol., 7 (6): 399–408, 1986. 16. J. V. D. Hough et al., Our experiences with hearing devices and a presentation of a new device, Ann. Otol. Rhinol. Laryngol., 95 (1): 60–65, 1986. 17. A. Tjellstrom et al., Osseointegration titanium implants in the temporal bone, Am. J. Otol., 2: 304–310, 1981. 18. B. Hakanson et al., The bone anchored hearing aid, Acta. Otol., 100: 229–239, 1985.

of patients with cochlear implants, Ann. NY. Acad. Sci., 311– 322, 1983. 24. D. K. Eddington and B. S. Wilson, Better speech recognition with cochlear implants, Nature, 352: 236–238, 1991. 25. W. F. Agnew and D. B. McCreery, Neural Prostheses: Fundamental Studies, Englewood Cliffs, NJ: Prentice-Hall, 1990. 26. D. K. Eddington, Speech discrimination in Deaf subjects with cochlear implants, J. Acoust. Soc. Am., 68 (3): 885–891, 1980. 27. B. S. Wilson, Speech processors for cochlear prostheses, Proc. IEEE, 76: 1143–1154, 1988. 28. P. J. Blamey, Acoustic parameters measured by formant estimating speech processor for multiple-channel cochlear implants, J. Acoust. Soc. Am., 82 (1): 38–47, 1987. 29. M. W. Skinner, Performance of postlingally deaf adults with the wearable speech processor (WSPIII) and mini speech processor (MSP) of the nucleus multi-electrode cochlear implant, Ear Hearing, 12 (1): 3–22, 1991. 30. H. McDermott, Anew portable sound processor for the University of Melbourne/Nucleus Limited multielectrode cochlear implant, J. Acoust. Soc. Am., 91 (6): 3367–3371, 1992. 31. A. M. Engebretson and M. French-St. George, Properties of an adaptive feedback equalization algorithm, J. Rehab. Res. Devel., 30: 8–16, 1993. 32. J. M. Kates, Feedback cancellation in hearing aids: Results from a computer simulation, IEEE Trans. Sig. Proc., 39: 553–562, 1991. 33. D .K. Bustamante and L. D. Braida, Principle-component amplitude compression for hearing impaired, J. Acoust. Soc. Amer., 82: 1227–1242, 1987. 34. H. Levitt and A. Neuman, Evaluation of orthogonal polynomial compression, J. Acoust. Soc. Amer., 90: 241–252, 1991. 35. P. M. Boers, Formant enhancement of speech for listeners with sensorineural hearing loss, Tech. rep., Inst. voor Perceptie Onderzoek, 1980. 36. H. T. Bunnell, On enhancement of spectral contrast in speech for hearing-impaired listeners, J. Acoust. Soc. Am., 82: 1227–1242, 1987. 37. W. Soede, A J. Berkhout, and F. A. Bilsen, Development of a directional hearing instrument based instrument based on array technology, J. Acoust. Soc. Amer., 94: 785–798, 1993. 38. W. Soede, F. A. Bilsen, and A. J. Berkhout, Assessment of a directional microphone array for hearing-impaired listeners, J. Acoust. Soc. Amer., 94: 799–808, 1993. 39. J. M. Kates and M. W. Weiss, A comparison of hearing-aid processing techniques, J. Acoust. Soc. Amer., 99: 3138–3148, 1996. 40. Cochlear Implants in Adults and Children, NIH Consensus Statement Online 1995, May 15–17, 13 (2): 1–30, 1995. http:// www.eaent.com/cochlear.html

19. B. Hakanson, A. Tjellstrom, and U. Rosenhall, Hearing thresholds with direct bone conduction versus conventional bone conduction, Scand. Audiol., 13: 3–13, 1984.

JAOUHAR MOUI¨NE

20. A. E. Geers, J. D. Miller, and C. Gustus, Vibrotactile stimulation: case study with a profoundly deaf child, 1983 Conv. Amer. SpeechLanguage-Hearing Assoc., Cincinnati, OH, abstract, p. 1, 1983.

Ecole Polytechnique de Montreal

21. D. Franklin, Wearable tactile aids: An alternative to cochlear implants, Voice, 1 (6): 4, 1985. 22. P. L. Brooks and B. J. Frost, Evaluation of a tactile vocoder for word recognition, J. Acoust. Soc. Am., 74 (1): 34–39, 1983. 23. B. J. Edgerton et al., The effect of signal processing by HouseUrban single-channel stimulator on auditory perception abilities

University of Sherbrooke

MOHAMAD SAWAN

HEART, ARTIFICIAL. See ARTIFICIAL HEARTS AND ORGANS. HEATING, ELECTRICAL. See INDUSTRIAL HEATING. HEATING RESISTANCE. See RESISTANCE HEATING. HEATING, SOLAR. See SOLAR HEATING.


}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP


Wiley Encyclopedia of Electrical and Electronics Engineering Human Motion Analysis Standard Article Ziad O. Abu-Faraj1, Gerald F. Harris1, Peter A. Smith1, Sahar Hassani1 1Shriners Hospital for Children—Chicago, Chicago, Illinois Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6606 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (652K)

●

●

● ●




Abstract The sections in this article are Evolution of Human Motion Analysis Anatomic Terminology The Cyclic Nature of Gait Stride and Temporal Gait Parameters Gait Analysis Methodologies Accuracy and Reliability of Motion Analysis Systems Biomechanical Modeling of Gait Data Ground Reaction Forces and Plantar Pressures Kinematics Kinetics


❍

❍ ❍



Dynamic Electromyography Energy Expenditure Interpretation and Decision Making Future Trends in Gait Analysis Acknowledgments Keywords: automated motion tracking system; biomechanical modeling; electromyography; EMG; energy expenditure; euler angles; gait analysis; gait cycle; ground reaction force; kinematics; kinetics; walking About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.


J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

HUMAN MOTION ANALYSIS Human locomotion is an acquired yet complex behavior requiring little thought during routine activities. It involves the integration of intricate signals from the nervous system, which result in muscle contraction and subsequent joint movement. Our comprehension of the development of coordinated gait activity, however, remains enigmatic despite numerous and profound advances in technology. The quest for knowledge in this field challenges various disciplines, such as biomechanical engineering, orthopedics, physical medicine and rehabilitation, kinesiology, physical therapy, and sports medicine. Gait analysis provides a quantitative measure of ambulatory activity. It is used to systematically assess joint kinematics and kinetics, dynamic electromyographic activity, and energy cost/consumption. Methods range from simple visual observation to video recordings and more sophisticated automated computer-based three-dimensional photogrammetric methods. Clinically, gait analysis has demonstrated effectiveness in pretreatment evaluation, surgical decision making, postoperative follow-up, and management of both adult and pediatric patients. It is also used as a tool to enhance elite athletic performance. Within the field of pediatric orthopedics, gait analysis has aided in advancing surgical treatment from single, isolated procedures to more comprehensive multilevel surgeries (1). These procedures include rectus femoris transfers and releases, hamstring lengthenings, tendoachilles lengthenings, gastrocnemius fascia lengthenings, osteotomies, and selective dorsal rhizotomies (2). Gait analysis has also proven useful in understanding more about orthopedic and neuromuscular disorders such as cerebral palsy, myelomeningocele, degenerative joint disease, multiple sclerosis, muscular dystrophy, rheumatoid arthritis, stroke, Parkinson disease, and poliomyelitis (1,2,3,4,5,6). Further applications include assessment of prosthetic joint replacement (7,8,9,10,11,12,13), analysis of athletic injuries (14,15), studies of amputation (16,17,18,19) and orthotic application (20,21,22), and the use of assistive devices (23). Additionally, gait analysis has been used in the evaluation of pharmacological treatment. Examples include botulinum toxin type-A (BOTOX manufactured by Allegran, Inc.), tetanus neurotoxin (TeNT), diazepam, baclofen, 45% ethanol, and phenol (24,25,26,27,28,29,30).

Evolution of Human Motion Analysis The study of locomotion is an ancient endeavor. Early documentation was demonstrated in the practice of Kung Fu by the Taoist priests in 1000 B.C. (31). An interest in locomotion was also manifested in the days of the ancient Greeks. Hippocrates (460 to 377 B.C.), in his work entitled, On Articulations, revealed interest in the relationship between motion and muscle function (31). Later, Aristotle (384 to 322 B.C.) investigated animal movement, which was depicted in his work Animal Spirits (32). Through a series of observations, Aristotle developed intuitive theories about the control of movement (33). It is believed that his grandson Erasistratus (310 to 250 B.C.), who was an anatomist and physician, was the first to discover the contractile property of muscle (32). Later, the Roman Empire witnessed the birth of the anatomical period. This era is attributed to the Greek physician Galen (131 to 210 A.D.), who worked for the Roman Emperor Marcus Aurelius (32). Galen 1

2

HUMAN MOTION ANALYSIS

classified exercise movement according to a paradigm that utilized the body segment, activity level, motion duration, and motion frequency (31). Thereafter progress remained marginal until the early seventeenth century with the foundation of modern motion principles by the Italian mathematician Galileo Galilei (1564 to 1642). Galileo described the time and distance parameters associated with moving objects (33). His work was later enhanced by that of the Italian mathematician, Giovanni Alfonso Borelli (1608 to 1679) (32,33). Borelli was the first to introduce mathematical principles to the study of motion, which was previously based on empirical observation. Borelli gave consideration to the motor force, the point of body support, and the resistance to be overcome (32). During the same period, René Descartes (1596 to 1650), a French mathematician, scientist, and philosopher, described the human body as a machine (33). His ideas were considered revolutionary in modern physiology. Both Borelli and Descartes theorized that all physiological processes obeyed the laws of physics (33). Sir Isaac Newton (1642 to 1727), quantitatively augmented the work of Galileo by introducing the concepts of dynamics, mass, momentum, force, and inertia (33). In 1687, Newton authored Principa, a publication in which he delineated the principles of dynamics using the three laws of motion and proposed the theory of universal gravitation (33). Paul J. Berthez (1734 to 1806) proposed a correlation between body force and muscle function (32). Around 1820, Chabrier demonstrated the relationship in muscle function between the free and fixed lower extremities (32). These contributions, while remarkable, remained purely observational. A new era in gait analysis emerged in 1836 in Leipzig with the work of the Weber brothers, Wilhelm and Eduard. With a combined background in physics, mathematics, anatomy, and physiology, the Weber brothers introduced the scientific foundations of the mechanics of human gait (32,33,34). Through a series of experiments, they formulated a mathematical model of the mechanics of human locomotion (32,33). They measured and reported on stance and swing phase, trunk movement, step duration, and step length (34). Even though their contribution was eminent, not all of their theories were accepted. The Weber brothers hypothesized that during swing phase, the limb advanced by gravitational force alone and required no muscular activity. Hence, they suggested that the swinging limb acts as a pendulum attached to the hip (32,33). This hypothesis was challenged and ultimately invalidated by Amand Duchenne (1806 to 1875). Duchenne, the Webers’ contemporary working in Paris, conducted a series of experiments with human subjects using electrical muscle stimulation (35). Duchenne demonstrated that it was impossible to advance the swinging limb in paralytic patients due to the absence of thigh flexors. Rather, these patients advanced their limbs by circumduction or by hip abduction. Hence, he concluded that circumduction was a compensatory mechanism, which would not be present if gravity alone was responsible for limb advancement (32,33). The first graphic plots of human gait were obtained at the Collège de France a` Paris and published in 1872 ´ by Gaston Carlet (1845 to 1892), a student of French physiologist and professor, Etienne Jules Marey (1830 to 1904) (36). Carlet was greatly influenced by the work of his mentor, who invented a special shoe-mounted measuring device that recorded foot pressure and contact duration during gait. This device consisted of an air compression chamber, constructed in the mid-metatarsal region of the shoe sole, and connected to a portable tambour and kymograph by means of rubber hoses. The tambour was mounted on the head to measure vertical oscillations of the body during gait. The recording cycle was triggered by pressure applied to a hand held squeeze ball and terminated with pressure cessation (37). The recording mechanism developed by Marey is illustrated in Fig. 1. Carlet modified Marey’s apparatus by adding a heel and forefoot chamber with a smoked drum mounted on an axis in the center of a 20 m circle. This design allowed extended and more detailed measurements (32,33). Figure 2 illustrates the apparatus used by Carlet for recording gait. The late 1800s witnessed a revolutionary advance in motion analysis along with the foundations of modern ´ cinematography. This success is primarily attributed to the creative work of two pioneers, Etienne Jules Marey (38) and Eadweard Muybridge (1830 to 1904) (39,40). Between 1872 and 1887 Muybridge conducted several photographic studies of human and animal locomotion. Figure 3 depicts one of Muybridge’s plates, which is a sequence of photographs of a walking child in the sagittal and coronal planes exposed simultaneously. In one of his early studies, Muybridge set up a network of wires across a horse track, and connected them to the shutters


3

´ J. Marey to study human locomotion. From Marey (37). Fig. 1. The recording mechanism developed by E.

of a linear array of still cameras. The running horse triggered the wires and sequentially exposed a series of still snapshots (33). In 1879, Muybridge invented the zoopraxiscope, a projector that reconstructed moving images from still photographs (41). The work of Muybridge was an inspiration to Marey, who recognized the importance of motion pictures in his studies of human walking. This realization led Marey to the invention of the chronophotograph, which was reported in 1885 (38). Marey also built a photographic “gun” to capture multiple images of a walking subject at adjustable intervals (33,34). Later, Marey refined his technique and used a black body suit with white reflective stripes to outline the body segments and obtain stick diagrams (33,34). In 1881, Karl Von Vierordt (1818 to 1884), a German physician and professor of medicine, contributed to the study of human kinematics by analyzing footprint patterns with colored fluid projections. This allowed a description of body segment movement in space during gait (34). In 1895, Wilhelm Braune and his apprentice, Otto Fischer, constructed a mathematical model of human gait consisting of 12 segments. They utilized cadaveric studies to determine anthropometric parameters, which included the center of gravity of each link segment and that of the total body. The model was capable of depicting displacements, velocities, accelerations, and forces during human gait (33,34,42). Braune and Fischer applied the methods of stereometry to the study of human motion. This novel methodology allowed tracking of the instantaneous location of a moving point in three-dimensional space. Consequently, they were able to obtain more accurate and precise calculations of kinematic parameters than those obtained from the two-dimensional photographs of their predecessors. Four cameras were utilized to capture the image of the moving subject. The coordinates were manually digitized

4


Fig. 2. The apparatus used by G. Carlet for recording gait. Reproduced with permission, from Schwartz and Heath (32).

Fig. 3. A sequence of photographs of a walking child. From one of Muybridge’s plates (40).

with the aid of a specially fabricated drafting table. In the experimental setup, Braune and Fischer utilized Geisler tubes mounted on a black jersey suit to outline body segments in a manner similar to that used by Marey. Each Geisler tube was filled with rarefied nitrogen and illuminated by an electrical current. Tubes were placed at the head, shoulders, wrists, hips, knees, and ankles. The joint center of each limb segment was identified by the end of the tube corresponding to that segment (33,42,43). Although it took from 6 h to 8 h to prepare a subject for the trial and several months to complete the analysis, the work of Braune and Fischer has been considered as a classical contribution (33,43). It was not until the twentieth century that a new approach to the analysis of human motion was intro¨ duced by Richard Scherb, Chief Surgeon at the Orthopaedic Institute Balgrist in Zurich, Switzerland. Scherb described the pattern and sequence of muscle action of the lower extremities during gait (43). In 1927, Scherb introduced a myokinesiographic method of recording muscle action during locomotion (33). In his early studies,


5

he utilized palpatory techniques to examine the onset and duration of muscle contractions in individuals walking on a treadmill (33,43). Scherb studied individuals demonstrating gait pathologies including poliomyelitis, spastic paralysis, and hemiplegia (43). In the same year, R. Plato Schwartz introduced the basograph, which was a new apparatus producing graphic records of plantar pressures in normal and pathological gait (44). Subsequently, Schwartz’s work underwent a series of refinements. In 1932 he utilized the pneumographic method of Carlet to determine alterations in plantar pressures during gait (32). Schwartz designed a pneumatic shoe to record the plantar pressures. The shoe consisted of three air compression chambers placed under the heel, fifth metatarsal head, and first toe. He was capable of obtaining continuous records of gait parameters by constructing a unique recording mechanism, consisting of pens made from acetate film and special capillary tubes. The experiments were conducted on a concrete floor and on a treadmill. During walking, the soles of the feet were outlined on a plate of ground glass via the use of a transillumination box, thus revealing the respective pressure points. Schwartz utilized a 16 mm motion camera to analyze his records. The period that followed World War II witnessed exceptional advances in the field of human motion analysis. During this period, the focus shifted toward the dynamic aspects of human gait. Vast contributions are attributed to Inman (1905 to 1980) at the University of California School of Medicine, San Francisco. In 1945, Inman joined Eberhart and Saunders in an extensive study of human locomotion. This project was a collaborative effort between the College of Engineering and the Medical School at the University of California at Berkeley. The work was designed to obtain quantitative data on normal human gait and that of amputees (45). The multidisciplinary nature of the project required the utilization of several different methodologies (33,43,45,46). A specially designed glass walkway, approximately 10 m long, was constructed for the study. Simultaneous exposure of displacements in the three anatomic planes was done by mounting mirrors beneath the glass surface. Velocities and accelerations of selected points were calculated with excellent precision from the displacement versus time measurements using graphonumerical differentiation. Sagittal displacements were acquired with interrupted light photography, following a similar technique to that designed by Marey. Transverse plane rotations of the lower extremities were obtained with a high degree of accuracy by surgically inserting pins at right angles into the bones of individuals. The degree of rotation at the pelvis, femur, and tibia were determined by measuring the relative displacements of the pins. Electromyographic techniques were also utilized to determine precisely the onset and cessation of dynamic activity of different muscle groups during the gait cycle. Additionally, a force plate dynamometer was utilized to determine the ground reactions: vertical force, torque, horizontal shear, and center of pressure (COP) on the foot. Other techniques were subsequently developed to determine the center of gravity locus, relative segment masses, and mass moments of inertia. Rotation measurements between the leg and foot required additional studies of cadaveric specimens. Results from this work were published in 1953 (46). In 1964, Murray (1925 to 1984) reported on a simple, economical, and repeatable method of recording gait (47). The purpose of this study was to obtain a database with baseline values for normal gait with which pathological gait could be compared. The work was completed in Milwaukee and was a result of a collaborative effort of the Wood Veterans Administration Hospital, the Marquette University School of Medicine, and the College of Engineering at Marquette University. The study resulted in the depiction of several kinematic gait parameters in the sagittal, coronal, and transverse planes. The parameters included the walking cycle duration and phases: stance phase, swing phase, and period of double limb support; step and stride lengths and widths; foot progression angles; sagittal rotation of the pelvis, hip, knee, and ankle; the vertical, lateral, and forward trajectories of the head and neck; the transverse rotation of the trunk and pelvis; and the sagittal excursions of the upper extremities. Sixty normal male subjects, separated into several categories based on age and height, participated in this study. The study employed interrupted-light photography in a low-light environment to capture the displacement patterns during gait. Subjects wore reflective fabric over selected anatomical segments and ambulated in front of a Speed Graphic camera with an open shutter. Markers were illuminated via an Ascor Speedlight stroboscope,

6


Fig. 4. Interrupted light photography. A typical photograph of a walking subject demonstrated by Murray in a study of human gait. Reproduced with permission, from Murray, Drought, and Kory (47).

flashing at a rate of 20 Hz. Additionally, an Ultrablitz flash unit was fired a single time during the walking trial, which allowed the identification of each marker position on the subject. The resultant photographs depicted a series of stick figures on a black background corresponding to the displacement of the reflective fabric with time. A mirror was mounted over the walkway to simultaneously capture transverse and sagittal plane images. Figure 4 depicts a typical photograph of a walking subject during one of the study trials. In 1972, Sutherland and Hagy presented a new method for the measurement of gait movements using motion picture technology (48). The novel method employed three 16-mm motion picture cameras positioned at the sides and the front of a walkway in an orthogonal fashion. Marked lines were placed at 30 cm intervals on the long axis of the walkway to provide distance and calibration references for the cameras. During subject testing, markers were placed over bony prominences of the anterior superior iliac spines, greater trochanters, knee joints, ankle joints, and dorsum of the feet between the second and third metatarsal heads. Following subject testing, the processed film was examined frame by frame on a Vanguard Motion Analyzer (Vanguard Instrument Corporation, New York). The projector featured a tilt control of the image, and a gear to control the X and Y coordinates with dial readout in thousandths of an inch. The film images of the walking subject were projected on a viewer, over which a digitizing grid was superimposed. Measurements were made directly from the projected images, and basic trigonometric functions were used to determine angular displacements. These measurements included transverse plane rotations of the pelvis, femur, and foot and sagittal plane rotation of the knee and ankle. The methods offered several advantages over those used


7

Fig. 5. The human body in the standard anatomical position and the three anatomical planes: sagittal, coronal and transverse.

previously. Subjects were not encumbered with an apparatus and data were recorded during a single session. The method was painless and did not cause discomfort. In addition, electromyograms could be recorded and superimposed onto the motion picture film. This technique also allowed simultaneous bilateral recording. The disadvantages included high equipment cost, and the amount of time required for analysis. Over the last few decades, increasingly accurate methods of studying human motion have evolved with more widespread clinical applications. Many contributors from various disciplines are responsible for these contributions. Continuous developments in the fields of microelectronics, instrumentation, and bioengineering have advanced the accuracy, reliability, and ease of use of current systems. Today, routine three-dimensional analysis includes joint angles, angular velocities, angular accelerations (kinematic analysis); ground reaction forces, joint forces, moments, and powers (kinetic analysis); and dynamic electromyographic activity (EMG analysis). Energy expenditure is also monitored during ambulatory testing in some laboratories.

Anatomic Terminology This section provides a narrative description of standard anatomic terminology and motion analysis parameters (49,50,51,52,53,54). Human locomotion is defined in all three anatomical planes: the sagittal, coronal, and transverse. These planes are commonly referenced to the human body in the standard anatomical position as depicted in Fig. 5. In this anatomic position, the individual is standing erect with the head facing forward, arms held at the sides, heels joined together, and the feet directed forward so that the great toes make contact.

8


Fig. 6. The types of motion in the three anatomical planes: (a) sagittal plane motion: pelvic tilt (anterior/posterior), hip flexion/extension, knee flexion/extension and ankle plantar flexion/dorsiflexion; (b) coronal plane motion: pelvic obliquity (up/down), hip abduction/adduction, knee valgus/varus and hindfoot valgus/varus; (c) transverse plane motions viewed from top: pelvic internal/external rotation, knee internal/external rotation, foot internal/external rotation and foot internal/external progression angle.


9

The sagittal plane is a vertical plane that separates the body or body segment into right and left sides. If the sagittal plane divides the human body exactly into left and right halves, then it is termed midsagittal plane. The midsagittal plane is also known as the median plane. The coronal (frontal) plane is defined as the vertical plane that separates the body or body segment into anterior (ventral) and posterior (dorsal) parts. Defined differently, the coronal plane is any vertical plane orthogonal to the sagittal plane. The transverse plane is a horizontal plane that separates the body or body segment into superior and inferior parts. In other words, the transverse plane is any plane orthogonal to both the sagittal and coronal planes. Directional expressions are used to describe the positions of the body or body segments in the three anatomical planes. The most commonly used directional expressions are anterior, posterior, superior, inferior, medial, lateral, proximal, and distal. Anterior (ventral) is the direction pointing toward the front of the body or body segment. Posterior (dorsal) is the direction pointing toward the back of the body or body segment. Superior (cephalic) is the direction pointing toward the head. Superior also denotes the upper part of a structure. Inferior (caudal) is the direction pointing away from the head toward the toes. It also refers to the bottom part of a structure. Medial is the direction that points toward the midsagittal plane of the body or midline of a structure. Lateral is the direction pointing away from the midsagittal plane of the body or midline of a structure. Proximal is the direction closer to the attachment of an extremity or limb to the trunk. Distal is the direction farther away from the attachment of an extremity or limb. Motion analysis requires that the movements of the body or body segments in the three anatomical planes be accurately described. Figure 6 illustrates the types of motion in the three anatomical planes: (a) sagittal plane motion; (b) coronal plane motion; and (c) transverse plane motion. Sagittal plane motion is often characterized by the terms flexion and extension. Flexion is described as the action that decreases the angle formed between two articulating bones, whereas extension is described as the action that increases the angle. Plantarflexion and dorsiflexion are expressions associated with foot and ankle motion. Plantarflexion is described as the excursion of the foot in the sagittal plane away from the anterior tibia. Dorsiflexion is the excursion of the foot in the sagittal plane toward the anterior tibia. The terms abduction, adduction, valgus, varus, inversion, and eversion are often associated with coronal plane motion. Abduction is defined as the action of moving a body segment away from the long axis, or midline, of the body in the coronal plane. Adduction is described as the motion of bringing the body segment back toward the midline of the body. Valgus is defined as the lateral angulation posture of the distal segment of a joint, whereas varus is defined as the medial angulation posture of the distal segment of a joint. Inversion and eversion are terms related to foot and ankle motion in the coronal plane. Inversion of the foot refers to the sole of the foot turning toward the midsagittal plane of the body, whereas eversion of the foot is the opposite motion. Motion in the transverse plane is typically restricted to joint rotations and foot progression angles. It is described in terms of internal and external directions. Using the right hand rule for right side joint rotations, if the fingers are curled in the direction of rotation, then internal rotation causes the thumb to point proximally. On the other hand, external rotation is the opposite gesture, causing the thumb to point distally. For left side joint rotations, the left hand rule is used. Additional terminology used in human motion analysis includes the center of mass (COM), center of gravity (COG), and center of pressure (COP) locations. These terms are often associated with force and moment measurements. The COM is an anatomical point used to locate the body segment mass in a global (threedimensional) reference coordinate system. The weighted average of the COM of each body segment represents the total body COM. The vertical projection of the COM on the ground is termed the center of gravity (COG). The center of pressure (COP) is the position of the vertical ground reaction force vector, and is the weighted average of all pressures acting on the plantar surface of the foot in contact with the ground. When one foot is in contact with the ground the resultant COP lies within that foot. When both feet are in contact with the ground, the resultant COP lies somewhere between the two feet. However, the exact location depends on the weight distribution between the two feet (55).

10


Fig. 7. The gait cycle and its eight events: initial contact (IC), loading response (LR), midstance (MST), terminal stance (TST), preswing (PSW), initial swing (ISW), midswing (MSW) and terminal swing (TSW).

The Cyclic Nature of Gait Human gait is a cyclic activity that can be described as a series of discrete events. The gait cycle is often defined as the period between initial contact of one foot with the ground and subsequent contact of the same foot (Fig. 7). The gait cycle consists of two major phases: stance phase and swing phase. Stance phase is that portion of the gait cycle when the foot is in contact with the ground, and typically represents approximately 62% of the total normal adult walking gait cycle (49,56,57). Three foot rockers (heel, ankle, and forefoot) occurring during stance phase, serve to control the forward fall of the body during normal ambulation. However, in pathological gait one or more of these rockers may not be present. Swing phase is defined as the period when the foot no longer contacts the ground and the limb advances in preparation for subsequent foot contact (49). Swing phase occupies the remaining 38% of the gait cycle. The gait cycle is also characterized by eight distinct events, which delineate in an orderly manner specific biomechanical functions (Fig. 7). Stance phase consists of five events: initial contact (IC), loading response (LR), midstance (MST), terminal stance (TST), and preswing (PSW). Swing phase, on the other hand, consists of the other three events: initial swing (ISW), midswing (MSW), and terminal swing (TSW) (56). Initial contact occurs when the foot strikes the ground and marks the beginning of stance phase. In normal walking, initial contact is often referred to as heel strike. For individuals with pathology, heel contact may not occur, hence the term IC is more appropriately used. During IC, the body COM is at its lowest position and the leg is positioned to begin stance with the heel rocker, also termed first foot rocker [Fig. 8(a)] (57,58). The heel rocker occurs as the heel contacts the ground at IC and progresses until the foot plantar flexes into full ground contact (foot flat). At IC, foot contact is made at a single point, causing the heel rocker to behave as an unstable lever system. Consequently, the foot is forced to pivot forward during the period of loading response (LR) with the fulcrum at the heel. During the heel rocker the pretibial muscles (tibialis anterior, extensor digitorum longus, extensor hallucis longus, peroneus tertius) undergo controlled eccentric (lengthening) contraction to resist the external moment created by gravity. This eccentric contraction causes the heel rocker to act as a shock absorber, which decelerates the foot at IC (57,58). Loading response is the first period of double limb support defined from IC (0%) to approximately 12% of the gait cycle (56,57). During this period, the limb acts as a shock absorber resulting in knee flexion, coincident


11

Fig. 8. The three foot rockers occurring during stance phase of the gait cycle: (a) the heel (first) rocker; (b) the ankle (second) rocker; (c) the forefoot (third) rocker.

with load acceptance and deceleration of the body. The period from IC through LR is termed weight acceptance. It is during weight acceptance that the leg provides weight-bearing stability, shock absorption, and control of forward progression (50,56,57). Single limb support marks the period from midstance (MST) through terminal stance (TST). During this period the opposite (contralateral) limb is in swing phase. Since normal walking is symmetrical this period occupies 38% of the total gait cycle. Midstance (MST) is the period immediately following the loading response (LR). It covers the first half of single limb support from approximately 12% to 31% of the overall gait cycle (50,56,57). Midstance begins when the contralateral foot clears the ground, initiating opposite limb swing phase, and ends at the instant when the body COM is decelerating as it passes over the stance limb forefoot. Midstance is the period of the ankle rocker (second foot rocker), when the ankle dorsiflexes [Fig. 8(b)] (57,58). Momentum forces the tibia to rotate forward over the plantigrade foot with the fulcrum at the ankle (57,58). The ankle rocker begins with foot flat and ends when muscle action restrains further dorsiflexion. This is caused by eccentric contraction of the plantar flexor muscles (gastrocnemius and soleus), primarily the soleus (57). Terminal stance (TST) comprises the second half of single limb support, which covers from 31% to 50% of the overall gait cycle (50,56,57). This event begins at the time of heel rise and extends until the contralateral limb contacts the ground (opposite foot contact). During this period, the body COM leads the forefoot and accelerates as it is falling forward toward the unsupported limb. Terminal stance is the period of the forefoot rocker (third foot rocker) [Fig. 8(c)] (57,58). This rocker begins at the end of MST and early TST as the body center of pressure (COP) approaches the metatarsal heads and the heel starts to rise. During this period, the metatarsophalangeal joints simulate a pivoted hinge, which functions as a rocker for the forward fall (57,58). This forward fall is initiated as the body COM leads the COP. During this period, the plantar flexors undergo concentric (shortening) contraction. This causes the forefoot rocker (third rocker) to serve as an acceleration rocker to prepare for limb advancement in PSW. Preswing concludes stance phase. It is also the final period of double limb support, which extends from 50% to 62% of the overall gait cycle (56). Preswing begins at IC of the contralateral limb and ends at terminal contact of the ipsilateral (stance) limb, just as the stance foot clears the ground (50,56,57). This period concludes stance phase and marks the beginning of swing phase. In normal gait, the hallux (great toe) is often the last

12


foot segment to clear the ground prior to swing. This final stance phase event is also termed “toe off.” For individuals with pathology, toe off may not occur, hence “foot off” becomes a more suitable term. Swing phase constitutes the last phase of the gait cycle, and is associated with limb advancement (56). During this phase, the swinging leg acts as a compound pendulum (56,59,60). The period of the pendulum is controlled by the mass moment of inertia. Variations in gait cadence are highly dependant on an individual’s ability to alter the period of this pendulum. Initial swing is initiated at toe off and progresses to the instant where the swinging limb is aligned with the contralateral limb. It is a period of modulated acceleration covering the time from 62% to 75% of the overall gait cycle and usually occupies one third of the swing phase (50,56). Midswing originates when the swinging limb is aligned with the contralateral limb, and is terminated when the swinging limb is in front of the stance limb (and the tibial shaft is vertical). It is a transitional period covering the middle third of swing phase from 75% to 87% of the overall gait cycle (50,56). Terminal swing is initiated with vertical tibial alignment and continues until IC. It constitutes the last third of the swing phase, from 87% to 100% of the overall gait cycle (50,56,57).

Stride and Temporal Gait Parameters The gait cycle is also characterized by stride and temporal parameters. These consist of step length (meters), step time (seconds), stride length (meters), stride time (seconds), walking speed (meters per second), cadence (steps per minute), single limb support time (seconds), double limb support time (seconds), and stance-to-swing ratio. These time and distance parameters provide an index of an individual’s walking patterns. Even though walking is a characteristic activity, there is slight variation in walking pattern from one individual to another. Any deviation in these parameters from normal values will challenge walking efficiency, and hence may affect energy expenditure (57). Step length is the longitudinal distance from IC of one foot to contralateral IC. Step time is the elapsed time associated with the step length. Stride length is the longitudinal distance between IC of one foot and subsequent ipsilateral IC. Normal gait is symmetrical, hence stride length is equal to twice the step length. Stride time is the elapsed time associated with the stride length. Walking speed is the rate of change of linear displacement along the predefined direction of progression per unit time. Cadence is defined as the rate at which an individual ambulates and is measured in steps per minute. The rate at which an individual ambulates at a self-selected comfortable speed is termed natural cadence. Single limb support is the elapsed time of the gait cycle during which one foot contacts the ground. Double limb support is the elapsed time of the gait cycle during which both feet are in contact with the ground. Single and double limb support may also be expressed as a percentage of the overall gait cycle. Stance-to-swing ratio is the stance interval divided by the swing interval (50,61,62,63). The stride and temporal gait parameters are highly dependant on one’s walking speed. Therefore, it is highly recommended that individuals walk at their freely selected cadence during a gait analysis exam. Although stride and temporal gait parameters are often helpful when diagnosing pathological conditions and evaluating treatment efficacy, they rarely provide sufficient insight into the origin of gait abnormalities (63).

Gait Analysis Methodologies Gait analysis has advanced a long way since the early days of Marey and Muybridge who utilized the photographic gun and multiple-still camera methods to describe human motion. Currently, several methods are employed in gait analysis to quantify this motion. These approaches include simple visual observation, video


13

recording, interrupted light photography, cinematography with manual digitization, electrogoniometry, multiaxial accelerometers, and automated three-dimensional motion tracking systems. Observational Gait Analysis. Observational gait analysis is a useful clinical tool. It is performed by simple visual observation of a walking individual. Although subjective, this method allows a trained examiner to identify many gait deviations during both stance and swing phases. This method is best performed by systematically focusing on one body part at a time. This work is often simplified with the aid of an evaluation form. Observing gait with the naked eye, however, is subject to numerous limitations. In this method, it is difficult to focus concurrently on multiple events and multiple body segments. According to Gage, events that occur faster than 1/16 s (62.5 ms) cannot be visually perceived (63). Hence, gait deviations may be missed even with a trained observer. Furthermore, this method cannot differentiate between primary abnormalities and compensatory responses. To avoid observational misinterpretation, discretion is advised. For example, apparent ankle equinus during initial contact may actually be a neutral ankle with flexed knee. Also, an apparent knee valgus during midstance may actually be a flexed knee and internal hip rotation. In spite of these drawbacks, observational gait analysis remains a useful clinical tool when used with other quantitative measures. Video Recording. The use of relatively inexpensive electronic equipment can provide refinement to observational gait analysis. A single video camera with a simple monitor (monochromatic color) and a video cassette recorder (VCR) yield a functional recording setup. The camera type can be based on a Vidicon tube or a charge-coupled-device (CCD) solid state detector. Today, video cassette recorders include advanced features such as freeze-frame, frame-by-frame view, and slow-motion replay. These features allow significant improvement over unaided visual observation. With this method, more consistent observations are obtained when motion videos are reviewed in slow motion rather than reviewing repeated normal speed walks. This method can be further expanded to accommodate simultaneous recording of sagittal and coronal plane motion. This can be simply done by adding a digital screen splitter and one or more cameras. There are varying opinions as to which planes of motion are most accurately analyzed with observational analysis. It is ideal to analyze pathology in the three anatomical planes (sagittal, coronal, and transverse). A limitation to this method is that it does not provide any information regarding dynamic electromyographic (EMG) activity or joint muscle torques, and data are not quantifiable. Cinematography with Manual Digitization. Motion picture or cine technology can be applied in gait analysis to film real-time events for analysis. This procedure is accurate yet time-consuming. The methods used are similar to those described by Sutherland et al. (48,61). Early investigators used markers mounted over anatomical landmarks and wooden wands affixed to pelvic and tibial belts to facilitate identification of body segment motion (61). Manual digitization of marker locations on a frame-to-frame basis allowed quantitative identification of marker positions in two-dimensional space with respect to the focal plane of the camera. This method can be extrapolated to three-dimensional space. However, the three-dimensional identification of marker positions requires the use of two or more cameras, since each marker must be seen by at least two cameras. Normally, walking is sampled at 50 to 60 frames per second (fps). On the other hand, when monitoring high-speed activities, higher sampling rates are needed. In these situations high-speed cameras with sampling rates of 200 fps or higher are used. The major disadvantage of cine with manual digitization is the overwhelming processing time needed for both data digitization and operator training (64). Interrupted Light Photography. Interrupted light photography is a simple technique that produces multiple images of a moving subject on a single photograph. This approach is similar to that employed by Murray (47). A subject, instrumented with reflective markers over anatomical landmarks, walks in front of a still camera with an open shutter to generate a series of stick figures. Interrupted light is produced by one of two methods. The first consists of a stroboscope firing at a rate of 30 flashes per second. The second method employs a rotating disk mounted in front of the lens and utilizes flood lights for illumination. The disk consists of several equispaced holes and rotates at a constant speed to generate the stroboscopic effect. This technique allows stride and temporal gait measurements directly from the photograph. Although restricted to

14


two-dimensional analysis, interrupted light photography is simple, inexpensive, and does not encumber the subject (33). Electrogoniometers. An electrogoniometer is a transducer that proportionally converts rotary motion to electrical current. It consists of two rigid links coupled by a potentiometer that measures the interposed angle. In measuring joint angular motion, the rigid links are strapped to a proximal and distal limb segment while the electrical output of the potentiometer marks the joint angle. Electrogoniometers can be applied to either two- or three-dimensional joint motion measurement (65,66). The design of electrogoniometers has been refined, allowing use in extremely difficult situations. Current electrogoniometric systems are more flexible and include designs such as parallelogram structures that allow movement outside the plane of measurement (33). Applications include knee joint motion analysis, where the instant center of rotation is continuously changing and cannot be accurately modeled as a simple hinge. Additionally, modifications to the basic electrogoniometer design have allowed its clinical use with orthotics and prosthetics (67,68). Even though electrogoniometer systems provide simple operation and real-time data measurement, they are difficult to apply, measure only relative joint angles, and can encumber a small individual (69). Multiaxial Accelerometers. Accelerometers are transducers used to measure linear and/or angular accelerations. They can be arranged in either uni- or multiaxial configurations. Displacement and velocity sensors can be used in combination with differentiator circuits to measure acceleration. Direct measurement of acceleration can also be obtained with the use of compact accelerometers. These devices are designed according to Newton’s second law and Hooke’s law (70). The measured acceleration may then be used to derive velocity and position data through numerical integration techniques. However, appropriate selection of initial conditions should be considered (71). Today, commercially available accelerometers allow the measurement of both linear and angular accelerations with six degrees-of-freedom. Automated Motion Tracking Systems. The most sophisticated method applied in human motion analysis employs automated tracking systems. These systems utilize two to seven cameras arranged around a calibrated capture volume. The cameras are positioned to cover motion in all three planes: sagittal, coronal, and transverse. The capture volume is a region in the laboratory space where the motion of interest occurs. The dimensions for the capture volume are based on demographic data and stride measurements. These values are obtained from published reports of pediatric and adult gait data. The two-dimensional images acquired from each camera are combined to obtain an instantaneous three-dimensional reconstruction of marker trajectories by using stereophotogrammetric techniques. The trajectories are usually described relative to a fixed laboratory frame system. Standard video technology allows sampling at 50 Hz or 60 Hz. However, some systems offer higher sampling rates for high-speed motion analysis. The sampling frequency in these systems can range from 200 Hz to 2,000 Hz. Currently, commercially available motion analysis systems provide unique marker and software packages and hardware characteristics. The specific analysis capability of each system relies on both the vendor-supplied hardware and software. Several of these systems offer optional features such as marker data filtering, generation of stick figures, analysis of joint velocities, and determination of joint moments and powers. User-friendly graphics aid in the presentation of the resulting data. Additionally, many of these systems provide user access to data files and establishment of a database. Marker Sets. Markers are used in conjunction with automated multicamera tracking systems. These markers are mounted over predetermined anatomical landmarks: bony prominences, joint axes, and limb axes. Spherical markers are often used, because they ensure that the centroid location of each marker is independent of the camera view angle. Transverse plane rotations typically use wands to increase measurement accuracy. Two types of markers are currently employed with these systems. The first type is a passive retroreflective marker. This type of marker is made of lightweight spheres covered with 3M 7610 reflective tape (St. Paul, Minnesota). Passive markers do not require power packs, but they do require a source of illumination. The reflected light is then captured by the cameras and digitized by the system. Light is usually supplied by strobes of light-emitting diodes (LEDs) arranged in one of two modalities: surrounding each camera lens or placed near


15

each camera. Flood lights may also be used to provide the source of illumination, but they are not recommended because they create visible distraction. An infrared light source is preferable to minimize subject distraction. In this configuration, cameras are equipped with optical filters, selective of light in the infrared spectrum (λ 860 nm) (72). Passive markers are useful in full-body gait analysis. However, in systems that are not fully automated user interaction is frequently required for marker identification. The second type of marker is an actively illuminated (optoelectric) marker, which is placed on the subject. In these systems, the light-emitting diode markers are pulsed at a predetermined frequency. This marker type allows higher sampling rates (200 Hz to 300 Hz), an increased number of markers per unit area, and frequency-coded data sorting. However, active markers require that the subject carry a power pack or tether, which may create subject distraction and gait alteration.

Accuracy and Reliability of Motion Analysis Systems Camera Positioning. Although the three-dimensional coordinates of a marker, whether active or passive, can be determined when viewed by two cameras, the realities of gait analysis necessitate that four to seven cameras be utilized. The objective of such a strategy is to obtain complete marker coverage at all times, preventing obstruction of one or more markers in situations such as arm swing and the use of assistive devices. If a marker is not viewed by at least two cameras concurrently, then the position must be estimated. A predictor corrector method is frequently used for marker position estimates. Marker dropout can significantly obscure joint motion data and, when manually supplemented, represents at best an educated estimate of the actual marker position. Such estimates can be deceptive, especially in the analysis of gait patterns (73). The use of multiple cameras increases the overlap and reduces marker dropout. More cameras are necessary to acquire bilateral gait data simultaneously. Marker Placement. In both active and passive marker systems, the overall accuracy of the system relies on optimal positioning of the markers with respect to anatomic landmarks. A major focus in marker set design is to maximize the distance between markers to reduce image overlap and sorting difficulties. Nonetheless, a resulting drawback is that small body segments such as children’s feet cannot always be completely identified or kinematically modeled. Guidelines for marker placement vary widely among systems. Marker placement depends on the biomechanical limb segment model and the procedure for determining joint centers utilized by the system. Common sources of error include inaccurate placement with respect to anatomical landmarks, skin and soft tissue movement, marker dropout from limb swing or assistive device obstruction, trunk rotation, and marker vibration (73,74,75,76). The estimated joint center locations and segment anthropometric data that are based upon markers can be utilized for a preanalysis snapshot, thus increasing the accuracy of characterizing the limb segment geometry with respect to the known marker locations. Figure 9 illustrates a common marker configuration used in lower extremity adult and pediatric gait analysis. Calibration and Linearization. To reduce potential sources of error, accurate methods of camera linearization and system calibration are required. Linearization is a process used to reduce inaccuracies inherent to camera architecture. These inaccuracies include lens geometry and optical distortion, uncertainty in focal length, aperture setting errors, deviations due to thermal strains, and other nonlinearities inherent in video scanning (77). A two-dimensional linearization grid consisting of a matrix of reflective discs located at precise coordinates is often used in the linearization process. The grid is affixed to a planar surface with proper alignment of the vertical and horizontal axes. The cameras are individually positioned with their focal plane parallel to the linearization grid. The objective is to obtain an image of the grid that covers the entire field of view of the camera. Once this is established, the perpendicular distance between the grid and the focal plane of the camera is measured. Subsequently, data is acquired and a linearization matrix is constructed to correct for errors. This matrix is virtually a one-to-one mapping from the measured target marker coordinates

16


Fig. 9. A common marker configuration used in lower extremity adult and pediatric gait analysis. In this configuration, surface markers are placed over the sacrum at middistance between the posterior superior iliac spines; anterior superior iliac spines; lateral condyles of the knee joint axes; lateral malleoli; and the dorsum of the feet between the first and second metatarsal shafts. Two markers (shown in black) are placed over the posterior heels during static trials only. Marker wands are placed laterally over the lower two-thirds of the femoral and tibial shafts.

to the true target marker coordinates. The linearization procedure should be conducted periodically. However, extensive use of the system requires that linearization be performed more frequently. Another important attribute in reducing potential errors is system calibration. This is used to correct for variations due to camera placement, temperature fluctuations, and sensor and electronics drift. System calibration is often described in terms of resolution and accuracy. Resolution describes the ability to discriminate position in terms of a linear measure and should be defined with reference to the laboratory capture volume. System accuracy quantifies the maximum absolute difference between a measured variable and its true value. System resolution is usually expressed in terms of millimeters, whereas system accuracy is often described in terms of a percentage of the known separation distance, usually the largest distance in the capture volume. Quantitative characterization of system resolution and accuracy is performed by placing markers at known positions in the capture volume. Measurements outside this capture volume are subjected to extrapolation errors (78). In most gait laboratory facilities, system calibration is routinely conducted on a daily basis. During subject testing both linearization and calibration files are retrieved to perform corrective measures on the kinematic and kinetic data.


17

Biomechanical Modeling of Gait Data In human motion analysis the marker set is coupled to a biomechanical or mathematical model that delineates the dynamics of various body segments (14,79,80,81,82). Kinematic data include the instantaneous linear and angular measurements of position, velocity, and acceleration. Measurements may be either absolute or relative. Absolute measurements are referenced with respect to a global (fixed) laboratory coordinate system. Usually, a cartesian coordinate system is selected with the origin located at a specific physical point in the laboratory. Relative measurements are depicted between neighboring body segments and require that the absolute orientation of each segment be obtained first. In three-dimensional motion analysis these relative measurements are commonly known as joint angles. Linear and angular velocities are obtained from the motion data by calculating the change in position per unit time. Typically, this is accomplished on a frameby-frame basis. The same method is applied for determining accelerations. In most gait analysis laboratories the convention is to describe the motion of the distal segment in relation to the next proximal segment. For example, in lower extremity gait analysis the foot is described with respect to the tibia, the tibia with respect to the thigh, the thigh with respect to the pelvis, and the pelvis with respect to the global laboratory coordinate system. Once the markers are identified in three-dimensional space, their collective position is used to describe the body segment motion characteristics. Biomechanical models are generally based on the assumption that body segments are rigid bodies. By definition, a rigid body is a system of mass points subject to holonomic constraints such that during motion a constant distance is maintained between all pairs of points (83). A rigid body in three-dimensional space must be represented by a minimum of three noncollinear markers (84,85,86). The spatial relationship among these markers describe the orientation, or attitude, of the rigid body in space. Subsequently, the position and spatial orientation of the rigid body is represented with six degrees-of-freedom or by six independent parameters (2). Motion that occurs in most anatomical joints is three-dimensional in nature. This motion consists of both translational and rotational components. The rigid body approach can be used in depicting joint motions by looking at the relative position of the proximal and distal body segments about the joint of interest. Each segment is represented by an independent imbedded coordinate system. The motion occurring at the joint is thus described in terms of the relative motion between the two embedded coordinate systems. To date, several methods have been employed in the description of three-dimensional joint motion. These methods include Euler (Cardan) angles (66,85,87), direction cosines (88), and finite helical (screw) axes (89,90). The Euler and direction cosines methods address only the rotational aspect of joint motion. The finite helical axes method describes the rotation of a rigid body about an axis defined in three-dimensional space, and a translation along that axis. In 1983, Grood and Suntay presented a unique method that utilizes a floating axis (87). Helical parameters have been used by Shiavi to describe knee joint motion (90). Seigler has also used helical parameters to describe motion at the ankle joint (91). The helical axis system, however, is difficult to interpret clinically and may be less useful for describing joint kinematics during gait (85). The Euler system is the most commonly used method clinically for describing three-dimensional motion. Euler angles describe a set of three successive finite rotations occurring in sequential order about predefined orthogonal (cartesian) coordinate axes. The order of rotations is critical and must be clearly defined. In gait analysis the standard order of rotation is first about the sagittal axis, then about the coronal axis, and finally about the transverse axis. The axis in question being perpendicular to the plane it represents. In this method, a fixed right-handed orthogonal coordinate system is established on each segment and moves with it. The motion of the distal segment is described relative to the next proximal segment. If a three-dimensional unit vector (i, j, k) is coupled to the x, y, z axes of the moving segment, and another unit vector (I, J, K) is coupled to the X, Y, Z axes of the fixed segment, then the relative orientation between the two segments after any arbitrary finite rotation may be described in terms of three Euler angles (α, β, γ), where α is the rotation about the sagittal axis of the fixed segment, β is the rotation about the line of nodes, and γ is the rotation

18


about the transverse axis of the moving segment. The line of nodes is a floating axis (coronal axis) orthogonal to both the sagittal axis of the fixed segment and the transverse axis of the moving segment. Each of these angles describes in sequence the rotation of the moving coordinate system with respect to the fixed (reference) coordinate system. In joint motion, the moving coordinate system is within the distal segment and the fixed coordinate system is within the proximal body segment or the laboratory global coordinate system. Figure 10 illustrates the application of Euler angles in describing the relative motion of the femur with respect to the pelvis. The Euler rotational matrix (92) is expressed as:

where c = cosine s = sine The fidelity of the biomechanical link segment model used in motion analysis relies on the accuracy of measurements and the reliability of several estimates. The model is coupled to the marker set and is subject to certain underlying assumptions about anthropometric characteristics, such as joint centers, segment length, segment masses, center of mass, and mass moments of inertia. For example, it is assumed that each greater trochanter (hip) marker is at a fixed distance from the center of hip joint rotation. With femoral deformity or hip anteversion this assumption is not accurate. Other sources of errors include any kinematic errors and errors in ground reaction force (GRF) measurement. Mathematical techniques used in the biomechanical gait models may also vary somewhat between facilities, thus the kinematic and kinetic data need to be interpreted carefully. This highlights the need for an accurate and thorough clinical assessment with any gait analysis.

Ground Reaction Forces and Plantar Pressures The ground reaction force (GRF) is an external force acting on the sole of the foot during the activities of standing, walking, or running. This force is a vectorial quantity and, therefore, has both a magnitude and a direction. The GRF is three-dimensional in nature and is usually resolved into a normal (vertical) component and two shear components: anterior-posterior and medial-lateral. Ground reaction forces are commonly measured with force plate dynamometers consisting of strain gage or piezoelectric transducers. In general, force plates are designed to acquired data from six channels, corresponding to the six degrees-of-freedom: Fx , Fy , Fz , M x , M y , and M z ; where F denotes the forces along the three global coordinate axes, and M represents the moments about these axes. Force plate data are typically sampled at 600 Hz (100 Hz/channel) and low-pass filtered at 10.5 Hz for quasistatic data and 1050 Hz for dynamic data. A sensitivity–calibration matrix is constructed exclusively for each force platform to convert between raw data (mV) and force (N) and moment (N·m) data. Processing of force platform output can provide GRF vector components including vertical load, anteriorposterior and medial-lateral shear loads, moments about the vertical axis, and location of the body COP. The vertical load pattern in normal individuals walking at their freely selected cadence follows an M-shaped curve with peak magnitudes of the order of 110% of body weight (80,93). Sutherland et al. reported that children exhibit somewhat diminished average vertical and shear peak forces as compared to adults (93). The vertical load curve is highly sensitive to any gesture that alters the ground reaction vector. For instance, the action of arm lifting can reduce the peak component to less than body weight (94). Additionally, both shear and COP measurements are influenced by the position and movement of all body segments, including the head, arms,


19

Fig. 10. Euler angle system utilizing the floating axis of Grood and Suntay (87). eα is the sagittal axis of the fixed segment (pelvis). eβ is the floating axis (coronal axis) orthogonal to both the sagittal axis of the fixed segment and the transverse axis of the moving segment. eγ is the transverse axis of the moving segment (femur).

trunk, pelvis, and legs. It has also been reported that disturbances such as pain, weakness, and unilateral hip pathology alter the vertical force pattern (94). Variations in cadence influence the magnitude and duration of the vertical load curve and have a direct effect on the gradient of the M curve, which indicates the rate of limb loading (95,96,97,98). Pressure measurements under the foot have also been of interest in human motion analysis. To date, numerous measurement techniques have been utilized in the study of the normal and pathological foot. These techniques include floor-mounted transducer matrices, pressure mats, instrumented shoes, insole-based pressure systems, and glass plates using the critical light reflection technique. Alexander et al. provided a recent review of the evolution of current foot-to-ground forces and plantar pressure measurement techniques and their clinical applications (99). Clinical studies of foot pressures have focused on the anaesthetic foot resulting from diabetes mellitus and Hansen’s disease (100,101,102), therapeutic footwear for the insensitive foot (103,104), the planovalgus foot in children with cerebral palsy (105,106), pedorthic inserts for the adult foot (107,108,109), distance running (110), and orthopedic walkers (111). Studies utilizing floor mounted transducers illustrate barefoot, isolated steps, and insole systems allow investigation of on-going step-to-step alterations in gait for longer durations. It should be noted that in plantar pressure studies consideration should be given to possible sources of error. These include sensor bending, hysteresis, nonlinearities, temperature and humidity changes, and stress shielding secondary to sensor-tissue or sensor-insole interface mechanics (109).

20


Kinematics Kinematics is the branch of engineering mechanics in which the motion of bodies is described without consideration of the underlying forces responsible for the movement (57,89). In human motion analysis, kinematics focuses on the study of the relative movement between body segments. These are frequently depicted as rigid link segments. Kinematic parameters include measurements of diarthrodial joint angles, displacements, velocities, and accelerations. In most clinical gait analysis reports, kinematic data are represented in the sagittal, coronal, and transverse planes. In the sagittal plane, joint angular motions include pelvic tilt, hip flexion/extension, knee flexion/extension, and ankle plantar flexion/dorsiflexion. In the coronal plane, joint angular motions consist of pelvic obliquity, hip abduction/adduction, and knee valgus/varus. Transverse plane joint angular motions include pelvic rotation, hip rotation, tibial rotation, foot rotation, and foot progression angle. The patient data are usually presented together with normal control data for comparison. Joint angle depictions may vary with different systems and are highly dependant on the marker arrangement and the biomechanical models employed (50). Furthermore, kinematic data can be supplemented with temporal and stride events that include cadence, walking speed, stride time, stride length, step time, step length, period of single limb support, and period of double limb support. A sample kinematic report is depicted in Fig. 11. Kinematic data are valuable in the analysis of gait disorders. However, they do not provide information on biomechanical efficiency (oxygen consumption and oxygen cost), ground reaction forces, joint moments, or joint powers. These measurements become important in circumstances where an ambulatory individual presents stable kinematic patterns, but reveals considerable variability in kinetic patterns (112). Furthermore, kinematic gait analysis of an individual with cerebral palsy may not reveal compensatory coping responses (57).

Kinetics Kinetics is the division of engineering mechanics in which motion is studied with consideration of the underlying forces that cause the movement. These forces include the external ground reaction forces and the internal joint, muscle, and ligamentous forces. The study of human motion analysis is governed by the application of Newton’s Second law and Euler’s equations of motion (113). The three-dimensional joint reaction forces and moments are obtained from both kinematic analysis and ground reaction forces. At each joint a state of equilibrium exists such that the internal joint reaction forces and moments balance the externally applied forces (114). Moments are often normalized to body weight and leg length, and are expressed as a percent of body weight times leg length (14). Joint powers are calculated once the moments, joint angles, and angular velocities are determined (115). The equations describing joint reaction forces are expressed in terms of Newton’s Second law as:

where Fx , Fy , Fz = are the sums of external forces acting on a limb segment in the x, y, and z directions, respectively, m = the mass of the limb segment, ax , ay , az = are the linear accelerations of the center of mass of the limb segment in the x, y, and z directions, respectively.


21

Fig. 11. A sample report of kinematic data in the three anatomical planes. The solid and dashed curves represent the patient’s left and right side motions, respectively. The dotted curves represent the mean motion of the laboratory normal sample group. The vertical lines separate stance phase from swing phase. Initial contact occurs at 0% of the gait cycle.

Joint moments are computed using Euler’s equations of motion as

22


where M x , M y , M z = are the sums of external moments applied to the limb segment in the x, y, and z directions, respectively, Ixx , Iyy , Izz = are the mass moments of inertia of the limb segment about the principal axes, αx , αy , αz = are the angular accelerations of the center of mass of the limb segment in the x, y, and z directions, respectively, ωx , ωy , ωz = are the angular velocities of the center of mass of the limb segment in the x, y, and z directions, respectively. Joint powers are calculated as

where Px , Py , Pz = are the joint powers in the x, y, and z directions, respectively, M x , M y , M z = are the external moments applied to the limb segment in the x, y, and z directions, respectively, ωx , ωy , ωz = are the angular velocities of the center of mass of the limb segment in the x, y, and z directions, respectively, P = the total joint power. In most clinical gait analysis reports, kinetic data includes hip, knee, and ankle joint moments and powers. Figure 12 illustrates a sample kinetic report. Clinically, kinetic analysis has proven useful in examining specific pathological conditions and operative procedures designed to restore normal function. Moment analysis has been useful for clinical decision making in cerebral palsy (116). It has also provided insight into subtle functional adaptations, such as the increased flexion moment observed at the hip and knee in patients with anterior cruciate ligament deficiency (117). Additionally, moment analysis has been recommended for predicting postoperative outcomes of high tibial osteotomy from preoperative knee adductor data (118). An area of increasing interest and expanding clinical application is that of biomechanical modeling, in which kinematic and kinetic gait data form an integral part of the solution. In contrast to the typical gait analysis solution, in which known motion and body segment parameters are used to estimate internal joint kinetics, these biomechanical models focus on an analysis of muscle function and effects. Typically, the more advanced biomechanical models address the following issues: alteration of muscle moment arm through surgery, alteration of muscle force generation through surgery, estimation of muscle length during normal and pathologic movement, and visualization and physical appreciation of interactions between muscle activity and kinetic gait parameters (119). Zajac et al. have developed biomechanical models based on detailed physiological considerations, including characteristics defined in the original Hill muscle model (120). Dynamics of contraction are determined by considering muscle output as two independent first-order processes: activation dynamics and contraction dynamics. A generic actuator is developed and scaled to specific muscle and tendon parameters including peak isometric force, optimal muscle fiber length, tendon slack length, and maximum shortening velocity. The model clearly demonstrates that the muscle acts as a central nervous system (CNS) controlled force generator, demonstrating a frequency response dependent on actuator length and/or activation level input. The muscle can also act as a spring, dashpot, or combined passive element, again dependent upon CNS control (120).


23

Fig. 12. A sample report of kinetic data in the three anatomical planes. The four rows represent sagittal plane internal joint moments; coronal plane internal joint moments; transverse plane internal joint moments; and total joint powers. The three columns represent: hip, knee, and ankle joint data. The solid and dashed curves represent the patient’s left and right side data, respectively. The dotted curves represent the mean results of the laboratory normal sample group. The vertical lines separate stance phase from swing phase. Initial contact occurs at 0% of the gait cycle.

Delp et al. have developed a graphics-based computer model to illustrate the effect of various surgical procedures upon gait kinematics and kinetics (119,121). These models use many of the physiological considerations included in Zajac’s models. Using a computer graphics work station, Delp et al. can alter the biomechanical models to study how various surgical procedures, such as osteotomies, tendon transfers, or tendon lengthenings, affect the moment arms and force-generating characteristics of the muscles. The models can also be used in conjunction with gait data to study muscle function during ambulation. Currently, the model represents an adult male with a height of approximately 1.8 m and a mass of 75 kg (119).

24


Dynamic Electromyography Dynamic EMG indicates muscle function by recording the voltage potentials generated by the electrochemical activities in the muscle. It provides detailed information about the timing and relative intensity of muscle activity. Nevertheless, dynamic EMG does not tell us about the strength of the muscle, whether the muscle is under voluntary control, or whether the contraction is isometric, concentric, or eccentric (122). The instrumentation needed for accurate recording of the EMG signal consists of recording electrodes, signal amplification and conditioning circuitry, signal transmission, and a means for data display and storage (123). Myoelectric signals can be recorded by three types of electrodes: needle, surface, and fine-wire. Needle electrodes are commonly employed for diagnosis of muscle disorders and are not recommended for gait analysis due to the discomfort they produce. In gait analysis, both surface and fine-wire electrodes are used for dynamic EMG analysis. Surface electrodes are placed on the skin over targeted muscles. Two types of surface electrodes are frequently used in routine gait analysis: Silver–silver chloride (Ag–AgCl) disks and active electrodes with built-in amplifiers and filters to improve signal quality and reduce noise effects. Silver–silver chloride electrodes consist of both a pickup and a reference disc. These electrodes require that the skin surface be cleansed and a conductive paste gel be used to minimize interface impedance at the recording site. It has been recommended that the Ag–AgCl disks be separated by 1 cm to obtain improved signal quality from the targeted muscle (124). Active electrodes have three terminals: reference, ground, and recording. These electrodes produce better signal quality because of their high input impedance and, hence, do not require epidermal preparation and conductive paste (124). Advantages of both types of surface electrode is that they are noninvasive, easy to apply, reusable, produce repeatable results, and can detect activities of muscle groups. However, disadvantages of both types of surface electrode is the inability to record activities from specific muscles and crosstalk from neighboring muscles. In situations where the monitoring of the activity of specific and deep muscles is required, the use of fine-wire EMG becomes necessary (125,126). Fine-wire EMG is a technique used to measure myoelectrical activity directly from individual muscles. In this method, a small-gauge hypodermic needle containing a pair of fine wire electrodes (one active, one reference) is inserted into the muscle of interest. The needle is then pulled out, leaving the two fine wires positioned within the muscle. Confirmation of the correct placement of the electrodes within the muscle is then determined by stimulating the muscle electrically through the fine-wire electrodes. Palpation and observation of the target muscle or tendon are used concurrently during the confirmation procedure (123,127). To avoid electric shorting, the electrodes are staggered by a few millimeters at their bared tips (123). The most common type of fine-wire electrode used in dynamic EMG is a nickel–chromium alloy wire (50 µm in diameter) with teflon insulation. The greatest advantage of fine-wire EMG is muscle selectivity. Recordings may be obtained from small peripheral muscles or those deeply located. However, this method is invasive, requires skilled placement, and may require multiple insertions. A new technique was introduced by Park and Harris in 1996 that avoids additional needle insertion. This technique monitors the electrical signal from the muscle while the needle with the fine-wire electrodes is advanced. This signal serves to guide the needle into the proper muscle (127). Muscle crosstalk is also present with fine-wire EMG. However the spectral content of the signal recorded with the fine-wire electrode allows filtering of some of the lower frequency volume conducted signals and thus allows reduction of muscle crosstalk. Both surface and fine-wire electromyographic signals have small amplitudes and do not allow direct interpretation without amplification. A differential amplifier with high common mode rejection (CMR) is used to eliminate electrical noise seen by both electrodes. The myoelectric signals recorded by surface and wire electrodes have different spectral characteristics of known bandwidths. Surface EMG signals have a bandwidth of 10 Hz to 350 Hz, with a mean frequency of 50 Hz. Fine-wire EMG signals have bandwidth of 10 Hz to 1000 Hz, with a mean frequency of 350 Hz. The lower bandwidth observed in surface EMG is a result of attenuation of higher frequencies as they travel through the tissues. The quality of the EMG signal is improved by filtering


25

Fig. 13. Dynamic EMG data. The first row represents a typical raw EMG signal recorded with active electrodes; the second row represents a full-wave rectification (absolute value) of the raw EMG signal; the third row represents the result of linear envelope (moving average) of the rectified EMG signal; and the fourth row represents the integrated processing (area under the curve) of the rectified EMG signal.

rejected frequencies. Motion artifacts, introduced by the subject or wires, have a low frequency content of 10 Hz to 20 Hz and can be reduced by increasing the low frequency cutoff. Commercial electromyographic systems are currently available with either cable or telemetry designs, or a combination of both. Cable systems are more reliable and less expensive than telemetric systems. However, they may encumber the subject with multiple tethers. Telemetric systems are based on radio frequency (RF) and are susceptible to electromagnetic interference. Telemetric systems may also require more frequent technical service. Other commercial systems are based on a combination of cable and telemetry design. These are capable of transmitting multiple signals on a single cable and offer numerous advantages over both systems. Automated methods for determining the onset and cessation of dynamic EMG have been reported (128). The raw EMG signal can either be analyzed or processed. A higher sampling rate is required to accurately acquire raw EMG signals. The most common methods of EMG signal processing are full wave rectification, linear envelope or moving average, and integration of the full wave rectified EMG (Fig. 13). The linear envelope is created by filtering a full-wave rectified signal with a low pass filter. The linear envelope is useful to assess on and off activity, but clonus bursts of muscle activity may not be seen (122). Many variables influence the recorded EMG signal such as magnitude of tension, velocity of shortening, rate of tension buildup, fatigue, and reflex activity (81). The relationship between the EMG signal and the force generated has been studied extensively (129,130,131,132,133,134) but needs to be interpreted with extreme caution in gait. Dynamic EMG does not indicate the strength of a muscle or the torque generation about a joint. There can be constant change throughout the gait cycle in multiple factors known to affect the relationship between the EMG signal and the force generated, such as the joint angle, the muscle fiber length, and the type of contraction (concentric, eccentric, or isometric). There is a great deal of interest in the relationship between the EMG signal and muscle force-joint torque. Early theoretical studies by Moore (135) and Libkind (136) suggested that during controlled isometric contraction, the EMG signal amplitude should increase as the square root of the muscle force generated. According to Basmajian and DeLuca, few experimental results

26


support the square root relationship (137). Muscle activity during gait is far more complex than simple isometric contraction, further complicating any direct relationship between the EMG signal and muscle force or joint torque. In pathologic gait, dynamic EMG is useful in preoperative evaluation of ankle and hip deformities (138, 139,140) and in analysis of rhizotomy results (141,142). Dynamic EMG in conjunction with kinematic analysis plays an important role in evaluating gait in individuals with neuromuscular disorders. The dynamic EMG signal is analyzed to identify whether the rectus femoris is firing continuously during swing phase. Posterior transfer of the distal rectus femoris has been found useful to augment knee flexion during swing in some individuals (143). Overall indications for rectus femoris transfer include a positive Duncan-Ely test, dynamic EMG evidence of prolonged swing phase rectus femoris activity, and reduction of swing phase knee motion by at least 20% (144,145). Analysis of dynamic EMG timing is also important in evaluating ankle valgus and varus deformities, although there is controversy regarding the role of dynamic EMG in posterior tibialis surgical procedures in cerebral palsy (133,140,146).

Energy Expenditure Gait requires kinetic energy expenditure as the body segments move, and potential energy generation as ligaments and elastic elements of muscle are stretched and as the COM moves vertically (53,57). Only 50% of potential energy is recovered during gait (147). An important role is played by two joint muscles that transfer energy between proximal and distal segments (148). The six determinants of gait, as described by Saunders, Inman, and Eberhart (46), work to minimize the total excursion of the body center of gravity and thus conserve kinetic energy. The body center of mass is highest during midstance, which is when the horizontal displacement is the least. The horizontal displacement is greatest during double limb support, which occurs when the COM and potential energy are lowest. Analysis of this two-dimensional sinusoidal pattern of displacement of the COM may underestimate the energy expenditure, because it may not properly reflect changes in body segment energy other than that of the trunk (57,149,150). The energy requirements for ambulation can be assessed during gait analysis. Heart rate data have been described as an index of energy expenditure for normal children and children with cerebral palsy (151). A linear relationship between heart rate and oxygen uptake at submaximal heart rates has been noted during normal gait (152). Information regarding oxygen uptake can be gathered with a modified Douglas bag technique or a mobile gas analysis system on a cart that is pushed alongside the subject during ambulation. Oxygen consumption (mL O2 /kg/min) and oxygen cost (mL O2 /kg/m) can be calculated. Heart rate and oxygen uptake data during ambulation are compared to data gathered during quiet sitting and quiet standing, and to agematched normals because there is an age-dependent linear relationship between walking velocity and oxygen consumption (153).

Interpretation and Decision Making The most frequent use of gait analysis is as a quantitative aid in surgical decision making. Other useful information is obtained through careful physical examination. Gait analysis has also proven beneficial in documenting the effects of treatment and to describe the natural progression and history of various neuromuscular conditions. The interpretation of quantitative gait data requires a multidisciplinary team with expertise in engineering, kinesiology, physical therapy, and medicine. The methods employed involve presentation of data in an understandable form and evaluation of past studies, if available, to define a natural history. In those


27

individuals with pathology a recommendation for intervention, if appropriate, is based upon available treatment modalities and technology. Patient data should be presented in a clear and understandable format. This requires a graphic representation of kinematic, kinetic, and EMG data, as well as that of a normal control population. In addition, review of photographs of subject marker placement, and video tapes of walking are helpful in understanding and appreciating the motion patterns. Radiographs are also reviewed at the time of analysis in those cases where musculoskeletal pathology is involved. Sutherland et al. have described the important effects of growth and maturation on the development of mature walking patterns in normal children, and this process is also known to occur over a longer time span of many years in children with neuromuscular disorders (154). Continued observation with maturation is necessary to track developmental abnormalities over time. The interpretation of gait analysis requires that deviations from normal be identified. These deviations are then separated into primary abnormalities and secondary compensations. A primary abnormality is defined as deviation from the normal gait pattern caused by contracture or muscle abnormality in an affected joint. Secondary compensations are strategies employed to optimize gait. For example, a person may walk with plantar flexion of the ankle because of a limb length discrepancy and shortening of the affected side. The examination of the patient along with EMG and kinetic data are useful in separating primary abnormalities from secondary compensations. Clinical recommendations are generally made to improve function. There are many ways to define a normal functioning gait pattern, including temporal and stride characteristics. More specifically, the prerequisites for normal gait can be classified into five areas: (a) stance phase stability; (b) adequate foot clearance; (c) preposition of the foot in swing; (d) adequate stride length; and (e) energy conservation through minimization of the excursion of the COG (57). Focusing on these specific prerequisites can aid the clinician in making treatment recommendations. The process of gait interpretation and decision making varies among laboratories. Most laboratories document the evaluation with a written report detailing gait abnormalities, and treatment recommendations with reference to any previous studies. The report is generally prepared by a multidisciplinary group, as described above. Most clinical gait laboratories also retain records of patients to comprise a database, which can be referenced for particular individual patients and also to study groups of patients sharing similar pathologies.

Future Trends in Gait Analysis Current and future directions in gait analysis will include more sophisticated tools for the analysis and interpretation of data such as pattern analysis, neural networks, and artificial intelligence. There is potential for much greater clinical application of moment and power data. Future biomechanical modeling of gait data will routinely include upper body segments and allow analysis of the flow of energy and power between body segments. More sophisticated models will also allow accurate analysis of the biomechanical effects of orthotics, prosthetics, and assistive devices. Because of competition and technological advances, it is expected that gait analysis systems will become less expensive and more accessible for routine clinical applications. Data banks with pretreatment and posttreatment results will need to be established through multicenter clinical studies. Continued mathematical synthesis of gait, anthropometric, and physiological data will improve musculoskeletal computer modeling, which in turn may improve pretreatment assessment, surgical planning, and postoperative followup.

28


Acknowledgments The authors would like to extend their appreciation to Mr. Gautam Sampath for his kind assistance in preparing this manuscript. The authors would also like to express their deep sense of gratitude to the Department of Graphic Arts at Shriners Hospital for Children—Chicago and in particular to Ms. Cynthia S. Armstrong Rosa for the art work. Much of the work referred to in this text was supported through the Shriners Hospitals for Children Research Foundation.

BIBLIOGRAPHY 1. P. A. DeLuca Gait analysis in the treatment of the ambulatory child with cerebral palsy, Clin. Orthopaed., 264: 65–75, 1991. 2. R. B. Davis P. A. DeLuca Clinical gait analysis: Current methods and future directions, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 17–42. 3. D. H. Sutherland Gait analysis in neuromuscular disease, San Diego Children’s Hospital Instructional Course, 1990, pp. 333–341. 4. S. J. Olney et al. Work and power in gait of stroke patients, Arch. Phys. Med. Rehab., 72: 309–314, 1991. 5. R. C. Wagenaar W. J. Beek Hemiplegic gait: A kinematic analysis using walking speed as a basis, J. Biomech., 25: 1007–1015, 1992. ˜ 6. R. B. Davis P. A. DeLuca S. Ounpuu Analysis of gait, in J. D. Bronzino (ed.), The Biomedical Engineering Handbook, Boca Raton, FL: CRC Press, and Piscataway, NJ: IEEE Press, 1995, pp. 381–390. 7. M. C. Collopy et al. Kinesiologic measurements of functional performance before and after geometric total knee replacement, Clin. Orthopaed. Rel. Res., 126: 196, 1977. 8. N. Rittman et al. Analysis of patterns of knee motion walking for four types of total knee implants, Clin. Orthopaed. Rel. Res., 155: 111–117, 1981. 9. M. P. Murray et al. Kinesiologic measurements of functional performance before and after double compartment marmor knee arthroplasty, Clin. Orthopaed. Rel. Res., 173: 191–199, 1983. 10. E. Olsson Gait analysis in hip and knee surgery, Scand. J. Rehab. Med., 15S: 5–53, 1986. 11. A. T. Berman et al. Quantitative gait analysis after unilateral or bilateral total knee replacement, J. Bone Joint Surg. 69: 1340–1345, 1987. 12. A. T. Berman et al. Quantitative gait analysis after unilateral or bilateral total hip replacement, Arch. Phys. Med. Rehab. 72: 190–194, 1991. 13. A. Wykman E. Olsson Walking ability after total hip replacement, J. Bone Joint Surg. 74B: 53–56, 1992. 14. T. P. Andriacchi R. P. Mikosz Musculoskeletal dynamics, locomotion, and clinical applications, in V. C. Mow and W. C. Hayes (eds.), Basic Orthopaedic Biomechanics, New York: Raven Press, 1991, pp. 51–92. 15. R. Jacobs G. Schenau Intermuscular coordination in a sprint pushoff. J. Biomech. 25 (9): 953–965, 1992. 16. R. L. Waters et al. Energy cost of walking of amputees: The influence of level of amputation, J. Bone Joint Surg. 58A: 42–46, 1976. 17. H. B. Skinner D. J. Effeney Gait analysis in amputees, Amer. J. Phys. Med., 64: 82–89, 1985. 18. A. Gitter J. M. Czerniecki D. M. DeGroot Biomechanical analysis of the influence of prosthetic feet on below-knee amputee walking, Amer. J. Phys. Med. Rehab., 70: 142–148, 1991. 19. G. R. Colborne et al. Analysis of mechanical and metabolic factors in the gait of congenital below knee amputees: A comparison of SACH and Seattle feet, Amer. J. Phys. Med. Rehab. 71: 272–278, 1992. 20. J. F. Lehmann et al. Gait abnormalities in peroneal nerve paralysis and their correction by orthoses: A biomechanical study, Arch. Phys. Med. Rehab. 67: 380–386, 1986. 21. J. F. Lehmann et al. Gait abnormalities in hemiplegia: Their correction by an ankle-foot orthosis, Arch. Phys. Med. Rehab. 68: 763–771, 1987. 22. D. S. Brodke et al. Effects of ankle-foot orthoses on the gait of children, J. Pediatr. Orthopaed., 9: 702–708, 1989.


29

23. C. Montecucco G. Schiavo Mechanism of action of tetanus and botulinum neurotoxins. Mol. Microbiol. 13 (1): 1–8, 1994. 24. L. Logan K. Byers-Hinkley C. D. Ciccone Anterior versus posterior walkers: A gait analysis study, Dev. Med. Child Neurol. 32: 1044–1048, 1990. 25. J. F. Mooney III L. A. Koman B. P. Smith Neuromuscular blockade in the management of cerebral palsy, Adv. Oper. Orthopaed., 1: 337–343, 1993. 26. L. A. Koman et al. Management of cerebral palsy with botulinum-A toxin: Preliminary investigation, J. Pediatr. Orthopaed. 13: 489–495, 1993. 27. A. L. Albright A. Cervi J. Singletary Intrathecal baclofen for spasticity in cerebral palsy, J. Amer. Med. Assoc., 265: 1418–1422, 1991. 28. E. B. Carpenter Role of nerve blocks in the foot and ankle in cerebral palsy: Therapeutic and diagnostic, Foot Ankle 4: 164–166, 1983. 29. E. B. Carpenter D. G. Seitz Intramuscular alcohol as an aid in management of spastic cerebral palsy, Dev. Med. Child Neurol. 22: 497–501, 1980. 30. G. Tardieu et al. Treatment of spasticity by injection of dilute alcohol at the motor point or by epidural route. Clinical extension of an experiment on the decerebrate cat, Dev. Med. Child Neurol. 10: 555–568, 1968. 31. J. V. Basmajian Therapeutic Exercise, 3rd ed., Baltimore: Williams & Wilkins, 1978, pp. 1, 3, 5. 32. R. P. Schwartz A. L. Heath The pneumographic method of recording gait, J. Bone Joint Surg. 14: 783–794, 1932. 33. M. E. Johanson Gait laboratory: Structure and data gathering, in J. Rose and J. G. Gamble (eds.), Human Walking, 2nd ed., Baltimore, Williams & Wilkins, 1994, pp. 201–224. 34. A. Steindler Kinesiology of the Human Body, Springfield, IL: Charles C Thomas, 1970, pp. 631, 632. ´ 35. G. B. A. Duchenne De l’Electrisation Localisée et son Application a` la Pathologie et a` la Thérapeutique, 3rd ed., Paris: Baillière JB et Fils, 1872. 36. G. Carlet Essai Expérimentale sur la Locomotion Humaine, Ann. Sci. Naturelles, Zool., 15: série 5, 1872. 37. M. Marey De la Locomotion Terrestre chez les Bipedès et les Quadrupèdes. J. Anat. Physiol. Homme Animaux 9: 42–80, 1873. ´ J. Marey La Méthode Graphique dans les Sciences Expérimentales, Paris: G. Masson, 1885. 38. E. 39. E. Muybridge Animals in Motion, 5th ed., London: Chapman & Hall, 1925. 40. E. Muybridge The Human Figure in Motion, New York: Dover, 1955. 41. E. Muybridge The New Encyclopaedia Britannica, 15th ed., Chicago, IL: Encyclopaedia Britannica, 1989, vol. 8, pp. 459–460. 42. C. W. Braune O. Fischer Der Gang des Menschen. 1. Theil. Versuche am Umbelasteten und Belasteten Menschen, Leipzig: S. Hirzel, 1895. 43. A. Steindler A historical review of the studies and investigations made in relation to human gait, J. Bone Joint Surg., 35A (3): 540–542, 728, 1953. 44. R. P. Schwartz W. Vaeth A method for making graphic records of normal and pathologic gaits; description of a new apparatus, the basograph. J. Amer. Med. Assoc. 90: 86, 1928. 45. H. D. Eberhart et al. Fundamental studies on human locomotion and other information relating to design of artificial limbs, Rep. National Res. Council, Committee Artif. Limbs, Berkeley: Univ. California, 1947. 46. J. B. Dec M. Saunders V. T. Inman H. D. Eberhart The major determinants in normal and pathological gait, J. Bone Joint Surg. 35A (3): 543–558, 1953. 47. M. P. Murray A. B. Drought R. C. Kory Walking patterns of normal men, J. Bone Joint Surg., 46A (2): 335–360, 1964. 48. D. H. Sutherland J. L. Hagy Measurement of gait movements from motion picture film, J. Bone Joint Surg. 54A (4): 787–797, 1972. 49. C. L. Vaughan B. L. Davis J. C. O’Connor The three-dimensional and cyclic nature of gait, in Dynamics of Human Gait. Champaign, IL: Human Kinetics, 1992, pp. 7–14. ˜ 50. S. Ounpuu Terminology for clinical gait analysis, Rep. Amer. Acad. Cerebral Palsy Developmental Medicine Gait Lab Committee, April 1994, Draft no. 2, pp. 1–20. 51. J. S. Thompson An introduction to gross anatomy, in Core Textbook of Anatomy, Philadelphia: J. B. Lippincott, 1977, pp. 1–13. 52. J. G. Hay J. G. Reid Introduction, in The Anatomical and Mechanical Bases of Human Motion, Englewood Cliffs, NJ: Prentice-Hall, 1982, pp. 5–15.

30


53. V. T. Inman H. J. Ralston F. Todd Kinematics, in Human Walking, Baltimore: Williams & Wilkins, 1981, pp. 22–61. 54. G. F. Harris J. J. Wertsch Procedures for gait analysis, Arch. Phys. Med. Rehab. 75: 216–225, 1994. 55. D. A. Winter Background and definitions, in Anatomy, Biomechanics and Control of Balance During Standing and Walking, Waterloo, Ontario: Waterloo Biomechanics, 1995, pp. 1–4. 56. J. Perry Phases of gait, in Gait Analysis: Normal and Pathological Function. Thorofare, NJ: SLACK, 1992, pp. 9–16. 57. J. R. Gage Normal gait, in Gait Analysis in Cerebral Palsy, London: Mac Keith Press, 1991, pp. 61–100. 58. J. Perry Basic functions, in Gait Analysis: Normal and Pathological Function. Thorofare, NJ: SLACK, 1992, pp. 19–47. 59. R. Hicks et al. Swing phase control with knee friction in juvenile amputees, J. Orthopaed. Res., 3: 198–201, 1985. 60. S. Tashman R. Hicks D. Jendrejczyk Evaluation of a prosthetic shank with variable inertial properties, Clin. Prosthetics Orthotics 9: 23–25, 1985. 61. D. H. Sutherland et al. Methods, in The Development of Mature Walking, London: Mac Keith Press, 1988, pp. 3–23. 62. J. Perry Glossary, in Gait Analysis: Normal and Pathological Function. Thorofare, NJ: SLACK, 1992, pp. 495–502. 63. J. R. Gage Basic measurement techniques, in Gait Analysis in Cerebral Palsy, London: Mac Keith Press, 1991, pp. 12–36. 64. D. H. Sutherland Gait Disorders in Childhood and Adolescence, Baltimore: Williams & Wilkins, 1984, pp. 1–13. 65. L. A. Lamoreux Kinematic measurements in the study of human walking, Bull. Prosthetics Res. 3: 10–15, 1971. 66. E. Y. S. Chao Justification of triaxial goniometer for the measurement of joint rotation, J. Biomech. 13: 989–1006, 1980. 67. E. N. Zuniga et al. Gait patterns in above-knee amputees, Arch. Phys. Med. Rehab., 53: 373–382, 1972. 68. R. E. Hannah J. B. Morrison Prosthetic alignment: Effect on gait of persons with below-knee amputations, Arch. Phys. Med. Rehab., 65: 159–162, 1984. ˜ 69. J. R. Gage S. Ounpuu Gait analysis in clinical practice, Sem. Orthopaed. 4 (2): 72–87, 1989. 70. M. R. Neuman Physical measurements, in J. D. Bronzino (ed.), The Biomedical Engineering Handbook, Boca Raton, FL: CRC Press, 1995, pp. 728–744. ˜ 71. R. B. Davis P. A. DeLuca S. Ounpuu Analysis of gait, in J. D. Bronzino (ed.) The Biomedical Engineering Handbook, Boca Raton, FL: CRC Press, 1995, pp. 381–390. 72. F. S. Abuzzahab et al. Foot and ankle motion analysis system: Instrumentation, calibration, and validation, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 152–166. 73. M. P. Kadaba et al. Repeatability of kinematic, kinetic, and electromyographic data in normal adult gait, J. Orthopaed. Res., 7: 849–860, 1989. 74. M. P. Kadaba H. K. Ramakrishnan M. E. Wootten Measurement of lower extremity kinematics during level walking, J. Orthopaed. Res., 8: 383–392, 1990. 75. R. Davis Sources of error in clinical gait analysis, in Clinical Decision Making in Gait Analysis, St. Paul, MN: Gillette Children’s Hospital (Course Syllabus) April 1992, pp. 177–179. 76. C. L. Vaughan M. D. Sussman Human gait: From clinical interpretation to computer simulation, in M. D. Grabiner (ed.) Current Issues in Biomechanics, Champaign, IL: Human Kinetics, 1993, pp. 53–68. 77. M. W. Whittle Calibration and performance of a 3-dimensional television system for kinematic analysis, J. Biomech., 15 (3): 185–196, 1982. 78. H. J. Woltring Planar control in multi-camera calibration for 3-D gait studies, J. Biomech. 13: 39–48, 1980. 79. D. A. Winter Kinematics, in Biomechanics and Motor Control of Human Movement, 2nd ed., New York: Wiley, 1990, pp. 11–50. 80. D. A. Winter Kinetics: Forces and moments of force, in Biomechanics and Motor Control of Human Movement, 2nd ed., New York: Wiley, 1990, pp. 75–102. 81. D. A. Winter Kinesiological electromyography, in Biomechanics and Motor Control of Human Movement, 2nd ed., New York: Wiley, 1990, pp. 191–212. 82. C. L. Vaughan J. H. Nashman M. S. Murr What is the normal function of tibialis posterior in human gait, in M. D. Sussman (ed.), The Diplegic Child, Rosemont, IL: American Academy Orthopaedic Surgeons, 1992, pp. 397–409. 83. H. Goldstein Classical Mechanics, 2nd ed., Reading, MA: Addison-Wesley, 1980. 84. E. J. Antonsson A three dimensional kinematic acquisition and intersegment dynamic analysis system for human motion, doctoral dissertation. Cambridge, MA: Massachusetts Institute of Technology, 1982.


31

85. H. K. Ramakrishnan M. P. Kadaba On the estimation of joint kinematics during gait. J. Biomech., 24 (10): 969–977, 1991. 86. L. Meirovich Methods of Analytical Dynamics, New York: McGraw-Hill, 1970, pp. 123–124. 87. E. S. Grood W. J. Suntay A joint coordinate system for the clinical description of three-dimensional motions: Application to the knee, J. Biomech. Eng., 105: 136–144, 1983. 88. I. H. Shames Engineering Mechanics, 2nd ed., Englewood Cliffs, NJ: Prentice-Hall, 1967. 89. H. J. Woltring R. Huiskes A. DeLange Finite centroid and helical axis estimation from noisy landmark measurement in the study of human joint kinematics, J. Biomech., 18: 379–389, 1985. 90. R. Shiavi et al. Helical motion of the knee: Kinematics of uninjured and injured kness during walking and pivoting, J. Biomech., 20: 653–665, 1987. 91. S. Seigler J. Chen C. D. Schenck The three dimensional kinematics and flexibility characteristics of the human ankle and subtalar joints—Part I: Kinematics. J. Biomech. Eng., 110: 364–373, 1988. 92. K. R. Kaufman D. H. Sutherland Future trends in human motion analysis, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 187– 215. 93. D. H. Sutherland et al. Force-plate values by age, in The Development of Mature Walking, London: Mac Keith Press, 1988, pp. 163–177. 94. J. Charnley R. Pusso The recording and the analysis of gait in relation to the surgery of the hip joint, Clin. Orthopaed. Rel. Res. 58: 153–164, 1968. 95. R. D. Crowinshield R. A. Brand R. C. Johnston The effects of walking velocity and age on hip kinematics and kinetics, Clin. Orthopaed. Rel. Res., 132: 140–144, 1978. 96. R. A. Mann J. Hagy Biomechanics of walking, running, and sprinting, Amer. J. Sports Med., 8 (5): 345–350, 1980. 97. S. R. Skinner The correlation between gait velocity and rate of lower extremity loading and unloading, Bull. Prosthetics Res., 18: 303–304, 1981. 98. R. W. Soames R. P. S. Richardson Stride length and cadence: Their influence on ground reaction forces during gait, in D. A. Winter et al. (eds.), Biomechanics IX-A, Champaign, IL: Human Kinetics, 1978, pp. 406–410. 99. I. J. Alexander E. Y. S. Chao K. A. Johnson The assessment of dynamic foot-to-ground contact forces and plantar pressure distribution: A review of the evolution of current techniques and clinical applications, Foot Ankle 11 (3): 152–167, 1990. 100. P. W. Brand Repetitive stress in the development of diabetic foot ulcers, in M. E. Levin and L. W. O’Neal (eds.), The Diabetic Foot, 4th ed., St. Louis: C. V. Mosby, 1988, pp. 83–90. 101. J. A. Birke D. S. Sims Plantar sensory threshold in the ulcerative foot, Leprosy Rev. 57: 261–267, 1986. 102. M. E. Levin Saving the diabetic foot, Med. Times 108 (5): 56–62, 1980. 103. J. H. Bauman J. P. G. Ling P. W. Brand Plantar pressures and trophic ulceration: An evaluation of footwear, J. Bone Joint Surg., 45B (4): 652–673, 1963. 104. P. S. Schaff P. R. Cavanagh Shoes for the insensitive foot: The effect of a “rocker bottom” shoe modification on plantar pressure distribution. Foot Ankle 11 (3): 129–140, 1990. 105. P. A. Smith G. F. Harris Z. O. Abu-Faraj Biomechanical evaluation of the planovalgus foot in cerebral palsy, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 370–386. 106. Z. O. Abu-Farj et al. A holter-type microprocessor-based rehabilitation instrument for acquisition and storage of plantar pressure data in children with cerebral palsy, IEEE Trans. Rehab. Eng. RE-4: 33–38, 1996. 107. A. H. Chang et al. Multistep measurement of plantar pressure alterations using metatarsal pads, Foot Ankle Int., 15 (12): 654–660, 1994. 108. Z. O. Abu-Faraj et al. Quantitative evaluation of plantar pressure alterations with metatarsal and scaphoid pads, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 387–406. 109. Z. O. Abu-Faraj et al. Evaluation of a rehabilitative pedorthic: Plantar pressure alterations with scaphoid pad application, IEEE Trans. Rehab. Eng., 4: 1–10, 1996. 110. P. R. Cavanagh M. A. Lafortune Ground reaction forces in distance running, J. Biomech., 13 (5): 397–406, 1980. 111. J. A. Birke D. A. Nawoczenski Orthopedic walkers: Effect on plantar pressures, Clin. Prosthetics Orthotics, 12 (2): 74–80, 1988.

32


112. D. A. Winter Kinematic and kinetic patterns in human gait: Variability and compensating facts, Human Movement Sci. 3: 51–76, 1984. 113. D. T. Greenwood Principles of Dynamics, Englewood Cliffs, NJ: Prentice-Hall, 1965. 114. A. Seireg R. J. Arvikar The prediction of muscular load sharing and joint forces in the lower extremities during walking, J. Biomech., 3: 51–61, 1975. 115. D. A. Winter Mechanical work, energy, and power, in Biomechanics and Motor Control of Human Movement, 2nd ed., New York: Wiley, 1990, pp. 103–139. 116. K. Lai K. N. Kuo T. P. Andriacchi Relationship between dynamic deformities and joint moments in cerebral palsy, J. Pediatr. Orthopaed., 8: 690–695, 1988. 117. T. P. Andriacchi G. M. Kramer G. C. Landon The biomechanics of running and knee injuries, in G. Finerman (ed.), Amer. Acad. Orthopaed. Surgeons, Symp. Sport Med., Knee, St. Louis: Mosby, 1985, pp. 23–32. 118. C. C. Prodromos T. P. Andriacchi J. O. Galante A relationship between gait and clinical changes following high tibial osteotomy, J. Bone Joint Surg., 67A: 1188–1194, 1985. 119. S. L. Delp Computer modeling and analysis of movement disabilities and their surgical corrections, in G. F. Harris and P. A. Smith (eds.), Human Motion Analysis: Current Applications and Future Directions, Piscataway, NJ: IEEE Press, 1996, pp. 114–132. 120. F. E. Zajac Muscle and tendon: Properties, models, scaling and application to biomechanics and motor control, in J. Bourne (ed.), CRC Critical Reviews in Biomedical Engineering, Boca Raton, FL: CRC Press, 1981, pp. 359–411. 121. S. L. Delp et al. An interactive graphics based model of the lower extremity to study orthopaedic surgical procedures, IEEE Trans. Biomed. Eng., 37: 757–765, 1990. 122. J. R. Gage EMG fundamentals and interpretation, in Clinical Decision Making in Gait Analysis, St. Paul, MN: Gillette Children’s Hospital (Course Syllabus), 1992, pp. 45–50. 123. J. Perry Dynamic electromyography, in Gait Analysis: Normal and Pathological Function, Thorofare, NJ: SLACK, 1992, pp. 381–411. 124. J. V. Basmajian C. J. DeLuca (eds.), Muscles Alive: Their Functions Revealed by Electromyography, 5th ed., Apparatus, detection, and recording techniques, Baltimore: Williams & Wilkins, 1985, pp. 19–64. 125. J. Perry C. S. Easterday D. J. Antonelli Surface versus intramuscular electrodes for electromyography of superficial and deep muscles, Phys. Ther. 61 (1): 7–15, 1981. 126. C. J. DeLuca R. Merletti Surface myoelectric signal cross-talk among muslces of the leg. Electroencephalogr. Clin. Neurophysiol., 69: 568–575, 1988. 127. T. A. Park G. F. Harris “Guided” intramuscular fine wire electrode placement: A new technique. Amer. J. Phys. Med. Rehab. 75: 232–234, 1996. 128. R. A. Bogey L. A. Barnes J. Perry Computer algorithms to characterize individual subject EMG profiles during gait, Arch. Phys. Med. Rehab., 73: 835–841, 1992. 129. V. T. Inman et al. Relation of human electromyogram to muscular tension, Electroencephalogr. Clin. Neurophysiol. 4: 187, 1952. 130. B. Bigland O. C. J. Lippold The relation between force, velocity, and integrated electrical activity in human muscles, J. Physiol. (Lond.), 123: 214–224, 1954. 131. J. R. Close E. D. Nickel F. N. Todd Motor-unit action-potential counts: Their significance in isometric and isotonic contractions, J. Bone Joint Surg. 42A: 1207–1222, 1960. 132. S. Bouisset EMG and muscle force in normal motor activity, in J. E. Desmedt (ed.), New Developments in Electromyography and Clinical Neurophysiology, Basel: Karger, 1972, pp. 547–583. 133. H. S. Milner-Brown R. B. Stein The relation between surface electromyogram and muscular force, J. Physiol., 46: 549–569, 1975. 134. J. P. Weir L. L. Wagne T. J. Housch Linearity and reliability of the IEMG v torque relationship for the forearm flexors and leg extensors, Am. J. Phys. Med. Rehab., 71: 283–287, 1992. 135. A. D. Moore Synthesized EMG waves and their implications, Am. J. Phys. Med. Rehab., 46: 1302–1316, 1967. 136. M. S. Libkind II, Modeling of interference bioelectrical activity, Biofizika, 13: 685–693, 1968. 137. J. V. Basmajian C. J. DeLuca (eds.), Muscles Alive: Their Functions Revealed by Electromyography, 5th ed., EMG signal amplitude and force, Baltimore: Williams & Wilkins, 1985, pp. 187–200. 138. J. Perry et al. Gait analysis of the triceps surae in cerebral palsy, J. Bone Joint Surg. 56A (3): 511–520, 1960.


33

139. J. Perry et al. Electromyography before and after surgery for hip deformity in children with cerebral palsy, J. Bone Joint Surg., 58A (2): 201–208, 1976. 140. J. Perry et al. Preoperative and postoperative dynamic electromyography as an aid in planning tendon transfers in children with cerebral palsy, J. Bone Joint Surg., 59A: 531–537, 1977. 141. L. D. Cahan et al. Intrumented gait analysis after selective dorsal rhizotomy, Dev. Med. Child Neurol., 32: 1037–1043, 1990. 142. C. L. Vaughan B. Berman W. J. Peacock Cerebral palsy and rhizotomy, J. Neurosurg. 74: 178–184, 1991. 143. J. R. Gage et al. Rectus femoris transfer to improve knee function of children with cerebral palsy, Dev. Med. Child Neurol. 29: 159–166, 1987. 144. J. R. Gage Knee dysfunction in cerebral palsy, in Clinical Gait Anal. Symp., Newington, CT: Newington Children’s Hospital (Course Syllabus) 1991, pp. 37–42. 145. J. R. Gage Distal hamstring lengthening/release and rectus femoris transfer, in M. D. Sussman (ed.), The Diplegic Child, Rosemont, IL: American Academy of Orthopaedic Surgeons, 1992, pp. 317–339. 146. M. J. Barnes J. H. Herring Combined split anterior tibial-tendon transfer and intramuscular lengthening of the posterior tibial tendon, J. Bone Joint Surg. 73A: 734–738, 1991. 147. G. Cochran A Primer of Orthopaedic Biomechanics, New York: Churchill Livingstone, 1982, pp. 269–293. 148. H. J. Yack D. A. Winter R. Wells Economy of two-joint muscle. In C. E. Cotton, et al. (eds.), Proc. Can. Soc. Biomech. 1988, pp. 180–181. 149. D. A. Winter A. O. Quanbury G. D. Reimer Analysis of instantaneous energy of normal gait, J. Biomech. 9: 253–257, 1976. 150. D. A. Winter A new definition of mechanical work done in human movement, J. Appl. Physiol. 46: 79–83, 1979. 151. J. Rose et al. Energy cost of walking in normal children and in those with cerebral palsy: Comparison of heart rate and oxygen uptake, J. Pediatr. Orthopaed. 9: 276–279, 1989. 152. P. Butler et al. The physiological cost index of walking for normal children and its use as an indicator of physical handicap, Dev. Med. Child Neurol. 26: 607–612, 1984. 153. R. L. Waters et al. Energy-speed relationship of walking: Standard tables. J. Orthopaed. Res. 6 (2): 215–222, 1988. 154. D. H. Sutherland L. Cooper S. Woo The development of mature gait, J. Bone Joint Surg. 62A: 336–353, 1980.

ZIAD O. ABU-FARAJ GERALD F. HARRIS PETER A. SMITH SAHAR HASSANI Shriners Hospital for Children—Chicago


}{{}}

●

HOME ●

ABOUT US ●

CONTACT US ●

HELP


Wiley Encyclopedia of Electrical and Electronics Engineering Prosthetic Power Supplies Standard Article Abdelouahab Djemouai1 and Mohamad Sawan2 1Ecole Polytechnique de Montréal, Montréal, Québec, Canada 2Ecole Polytechnique de Montréal, Montréal, Québec, Canada Copyright © 1999 by John Wiley & Sons, Inc. All rights reserved. DOI: 10.1002/047134608X.W6605 Article Online Posting Date: December 27, 1999 Abstract | Full Text: HTML PDF (285K)

●

●

● ●



Search this title Enter words or phrases ❍

❍ ❍


Abstract The sections in this article are Basic Principle of Prosthetic Power Supplies Radio-Frequency Power Transfer Using an Inductively Coupled Link Main Inductively Coupled Link Approaches Summary and Future Developments Keywords: biomedical implanted electronic devices; biomedical prosthetics; biosensors; transcutaneous inductive links; inductively coupled links; radio-frequency (RF) power transmitter; power transfer efficiency; class D and E power amplifiers; voltage regulator; L-C resonant circuits About Wiley InterScience | About Wiley | Privacy | Terms & Conditions Copyright © 1999-2008John Wiley & Sons, Inc. All Rights Reserved.


J. Webster (ed.), Wiley Encyclopedia of Electrical and Electronics Engineering c 1999 John Wiley & Sons, Inc. Copyright

PROSTHETIC POWER SUPPLIES Sensors and actuators operating inside the body to recuperate human organs have generated a great deal of interest during the last two decades (1,2,3,4,5). Advances in microelectronics have led to the development of powerful miniaturized devices, among them the various implantable artificial organs, sensors, and electrical stimulators (6,7,8,9,10,11). These devices have recently been intensively used in many disease circumstances, for example, cardiac pacemakers, cochlear implants, bladder controllers, bone growth stimulators, phrenic nerve stimulators, and functional neural systems, which are dedicated to lower-extremity and upper-extremity movements. Most of these advanced, commercially available devices are extracorporally powered through radio-frequency (RF) transmitting energy systems (12,13). All of them necessitate external, high-efficiency sources of energy. The majority of available prosthetic power supplies are based on amplitude modulation (AM) transmission, which rectifies a high-frequency signal (a carrier) to power on the implantable part.

Basic Principle of Prosthetic Power Supplies Numerous alternatives are available for supplying power to implanted devices. The most important of these are the percutaneous plugs that break the skin to reach the implanted prosthesis. This connection may result in infection, risking implant damage and the safety of the patient. The second alternative is to use a battery (14). In this case, there are several disadvantages: frequent recharging or reimplantation is required and the battery occupies an important place relative to the device. The remaining alternatives are based on wireless techniques and they do not have any of the limitations mentioned previously. The goal of the transcutaneous energy transfer is to develop a safe and effective method for the remote delivery of energy to implanted biomedical devices, since this method draws power from an extracorporal source of energy and leaves the skin intact. These technologies allow patient mobility, improve quality of life, and reduce risk of infection. The transcutaneous links can be based on one of three categories of energy-transfer techniques: (1) electromagnetic, (2) optic (various forms of light), and (3) ultrasound. The electromagnetic radio-frequency inductively coupled link is the most commonly used technique to deliver power to prosthetics, and it is the most desirable method for patients. The inductively coupled power-transfer technique, however, lacks some characteristics such as high-energy-transfer efficiency and wide bandwidth, as well as transversal, orthogonal, and other inductance alignments. In order to improve most of these limitations, other transdermal powering techniques have been proposed. In fact, a transcutaneous optical power converter has been introduced in Ref. 15. This method consists of solar cells receiving light from an external light source, which can be either a halogen-lamp- or a laser-diodeilluminated skin surface. Neither the optical nor the ultrasound transcutaneous power-supply techniques give the expected performances and they have been employed only in very restricted cases. The remainder of this article focuses on the characterization of the inductively coupled link, which remains the most appropriate choice because of its greater safety, efficiency and high tolerance to displacement. Thus, the external controller transmits power indefinitely at field levels that meet government safety standards (10 mW/cm2 ) (8,12,16). 1

2

PROSTHETIC POWER SUPPLIES

Fig. 1. Block diagram of a prosthetic device using an inductively coupled link. The purpose of the link is to transfer power and data to the implant.

Radio-Frequency Power Transfer Using an Inductively Coupled Link A simplified generic block diagram of a system dedicated to power electronics implants using an inductive link is given in Fig. 1. The external part of such a system is driven by a power amplifier, which is the last stage of the RF transmitter. The implantable electronics device receives the necessary energy through this inductive link, which is followed by a rectifier and a voltage regulator. The voltage regulator delivers an adequate power supply and dc voltage to the electronics implant. The inductive link itself is composed of two series resonant circuits, a primary circuit (Rp , Cp , Lp ) and a secondary circuit (Rs , Cs , Ls ). Analysis of the Inductively Coupled Link. In designing inductively coupled links to transfer energy to implantable devices, the most important characteristics to be considered are the ratio of the output voltage to the input voltage, the power efficiency, the bandwidth, the external power amplifier, and the load of the implant. The ac output voltage of the internal inductively coupled link should be high in order to allow the voltage regulator to extract the expected dc voltage as well as the energy to power on the whole implant device (Fig. 1). Based on the constraints mentioned earlier, the remaining sections of this article are devoted to determining the characteristics of the inductively coupled link. We begin by studying an unloaded inductively coupled link similar to that given in Fig. 2, and then we study the same link when the load and the amplifier output impedance are taken into account. Ratio of the Output Voltage to the Input Voltage. The mesh equations of the unloaded inductively coupled link in Fig. 2 are the following:


3

Fig. 2. Unloaded inductively coupled link.

Fig. 3. Variation of the ratio (V out /V in ) of the coupled link with frequency for different coefficients of coupling k for the case where the primary and secondary circuits are tuned to the same resonant frequency f 0 .

is the mutual inductance between Lp and Ls , and k is the coefficient of coupling. Zp and Zs are the series impedances of the primary and secondary circuits, and they are given by

Since the coupled link is operated, in general, at the resonance frequency, let us assume that the primary (Rp , Lp , Cp ) and the secondary (Rs , Ls , Cs ) circuits are tuned to the same frequency f 0 , so that

4


By rearranging Eqs. (1a) and (1b), we can obtain the ratio of the voltage V out developed across the capacitor Cs of the secondary circuit (V out ) to the input voltage applied in series with the primary circuit (V in ), as described in Ref. 17:

where f is the operating frequency and Qp and Qs are the quality factors of Lp and Ls , respectively. The variation of |V out /V in | with the frequency for different values of the coupling coefficient k is presented in Fig. 3. At the resonance frequency f 0 , γ is equal to 1 and Eq. (4b) becomes

Figure 4 shows the variation of the ratio |V out /V in | as a function of k for a specific inductively coupled link. The maximum ratio is reached when the coefficient of coupling (k) is equal to the critical or optimal coupling kc , which is given by

The ratio V out /V in is affected by the reflected impedance Zref from the secondary (receiving coil) circuit to the primary coil. This impedance is defined as follows:

The effect of this impedance on the primary circuit is exactly as if an impedance equal to (ωM)2 /Zs had been added in series with the primary circuit. At resonance and at critical coupling, we have (1) and

(2)


5

Fig. 4. Variation of the ratio (V out /V in ) of the link with the coupling factor k at the resonance frequency f 0 . The maximum of this ratio occurs at k equal to the critical coupling factor kc .

The combination of Eq. (8b) with Eqs. (4b) and (4c) results in (3) then (4) Thus, the reflected impedance Zref is purely resistive and equal to the primary resistance Rp . With these conditions, the secondary current is maximal and the corresponding maximal possible voltage ratio V out /V in can be obtained from Eqs. (5) and (8b).

In this case, the corresponding dissipated power in the primary circuit can be calculated as follows:

Power Efficiency η l . Based on the fact that the power delivered to the reflected impedance Zref is equivalent to the same quantity of power transferred to the secondary (I2 s Zs ), the power efficiency η l of the

6


Fig. 5. Variation of power efficiency of the coupled link with the coupling factor k. The optimal value of this efficiency is equal to 0.5 which corresponds to the critical coupling factor kc .

inductively coupled link can be expressed as (18,19)

At the resonance frequency f 0 , Eq. (11) can be simplified to obtain

The variation in the power efficiency ηl with the coupling coefficient kc is depicted in Fig. 5. The optimum value of ηl is reached when both the secondary current Is and the output voltage V out are at their maximum values. This is possible only at the resonance frequency f 0 and at the critical coupling factor kc . Then, from the Eq. (12), the optimal efficiency ηlopt is

Thus, the optimal efficiency of an unloaded coupled link is equal to 50%, and this can be reached only at both the critical coupling factor and the resonance frequency. Bandwidth of a Tuned Inductively Coupled Link at the Critical Coupling Factor. The bandwidth of a tuned inductively coupled link at the critical coupling coefficient and with identical quality factor coils (Qp , Qs ) can be derived easily from the equations previously developed. In fact, at k = kc and Qp = Qs = Q, Eq. (4)


7

becomes

At the resonance frequency f 0 , and with γ = 1 and h = 0, Eq. (14) becomes

Equations (11) and (15) give

Since the operating frequency of the inductive link is usually√too close to the resonance frequency, and γ is so nearly unity between half-power points, thus to get A/Ar = 1/ 2, the condition

should be fulfilled. If we set f = f 0 + f /2, then Eq. (17) becomes

Generally f /f 0 1, so Eq. (18) can be simplified to lead to

Inductive Link with the Amplifier and Load Impedances. If the effective resistance at the input of the regulator is Rdc , and with the assumption that the rectifier is ideal, then according to Ko, Liang, and Fung (18), the voltage peak value across Cs appears as a direct voltage across Rdc (Fig. 1). The equivalent ac load resistor Rac , which dissipates an amount of ac power equivalent to the dc power in Rdc (16,18), is

Based on this assumption, a circuit equivalent to that in Fig. 1 can be obtained, as shown in Fig. 6(a). The resistor Rac acts as a load to the inductively coupled link where the energy should be transferred. Now, if this

8


circuit is working near the resonance frequency and if Rac is relatively high in comparison with Rs and ωLs , then this circuit can be transformed to obtain the circuit in Fig. 6(b). Here RL is the ac series resistor equivalent to the load resistance Rac (18) and is equal to

The circuit in Fig. 6(b) is similar to that in Fig. 2, except that the series resistors of the primary and secondary circuits are affected by the presence of the output resistance of the power amplifier (Rg ) and the resistance equivalent to the implant or load (Rac ). With these transformations, the equivalent series resistors (Rep , Res ) and the quality factors (Qep , Qes ) of the primary and secondary circuits, respectively, are given by

Replacing (Qp , Qs ) in Eq. (5) by (Qep , Qes ), we obtain the new transfer function at resonance frequency V out /V in :

The maximum of V out /V in occurs at the critical coupling k = kec , which is defined by

hence

The reflected impedance at resonance frequency is now given by

Zref can be expressed in terms of Rac , Rs , Qp and Qs , as (18)


9

Fig. 6. Inductively coupled link: (a) with the amplifier and load resistances and (b) with the equivalent series resistor of the load.

and at k = kec , Zref becomes equal to Rep , and at these conditions the power dissipated in the primary circuit is

The global power efficiency (η = Pout /Pin ) can be determined by the power efficiency of the inductive link ηl and the efficiency of the secondary circuit ηs , where

and

which could be expressed in term of Rac , Rs , k, Qp , and Qs , as follows:

Coupling Coefficient and Mutual Inductance. The coupling coefficient k is related to the mutual √ M inductance by the relation k = M/ L1 L2 , where L1 and L2 are the values of the primary and secondary inductances, respectively. Since the operating frequency of transcutaneous inductive links used to transfer power to prosthetics is in the radio-frequency range, the primary and secondary coils L1 and L2 have, in general, very small values. It follows that each of these coils is constructed of only one or a few small metal loops. For the general case, let us consider a system in which the primary and secondary coils are formed of one-turn circular loop each and are perfectly aligned (Fig. 7). The two coils are separated by the distance d and have r1 and r2 as radii.

10


Fig. 7. Illustration of two perfectly aligned coils.

The general form of mutual inductance M between two coils of n1 and n2 turns is given by the following Neuman relationship (20):

Using this relation, the mutual inductance M of the system in Fig. 7 is

where µ0 is the permeability of free space in henrys per meter and f is a variable defined by

and K(f ) and E(f ) are the complete elliptic integrals of the first and second kind (20), respectively, defined as

The values of L1 and L2 can be derived from Eq. (34), and are given by


11

Fig. 8. Variation of the coupling coefficient k with the coil separation. The coupling factor is high when the separation of the two coils is small and decreases as this separation increases.

Here ei is the wire radius of the coils. Using Eqs. (1) and (5), the variation of the coupling coefficient k as a function of coil separation d is shown in Fig. 8. The coupling coefficient is almost equal to 1 when the coil separation d is too small, but decreases rapidly as the separation d increases. When the primary and secondary coils present a lateral or an angular misalignment, the calculation of the mutual inductance is complex and there is no single and precise formula for M for these cases. A good discussion and approximation formulas for the mutual inductance, for the case of the presence of the different misalignments, can be found in 21 and 22. When the primary and secondary coils are constructed of more than one turn, a better arrangement of the coil loops described in Refs. 23 and 24 is preferred in order to enhance the coupling coefficient. In this arrangement, spiral coils are used in such a way that the turns are not concentrated at the circumferences, but distributed across the diameter. And, in order to reduce the magnetic flux leakage in the body, a set of amorphous fibers is attached radially on these spiral coils (23). These amorphous fibers increase the coupling coefficient and the inductance, which in turn reduces the magnetic flux leakage. Power Amplifiers. The inductive link dedicated to transferring power to prosthesis necessitates the employment of a high-efficiency power amplifier. This helps to transfer the energy required by the implant with minimum loss in the side of the transmitter. The drives most used in the transmitter of a system using an inductive link are either class D or class E amplifiers. The reason behind the choice of these classes of amplifiers is their switching mode in operation, which provides them with high power efficiency. The Class D Power Amplifier. An example of a class D amplifier is shown in Fig. 9. It uses a pair of power metal oxide semiconductor (MOS) transistors followed by a series resonant circuit (the primary circuit of the inductive link). Each transistor acts as a low-loss switch, which has only two states: the ON state (conducting) and the OFF state (nonconducting). The two transistors are alternately opened and closed at very high speed, resulting in a rectangular voltage waveform at the output of the amplifier. In its ON state, the MOS transistor acts as a small resistor; therefore the power loss is very small; in its OFF state, the leakage current is very small and again the power loss is negligible. When the resonant circuit is tuned to the

12


Fig. 9. Class D power amplifier based on a two MOS transistors (switches). The switch mode operation of these transistors helps to reduce the power loss of the amplifier. The amplifier works efficiently when operated at a frequency equal to the resonant frequency f 0 of the inductively coupled link.

Fig. 10. Class E power amplifier based on a single MOS transistor acting as a switch. As in the case of the class D amplifier, the switching mode operation of the transistor makes this kind of amplifier provide a very high power efficiency.

switching frequency of the transistors, it presents low reactance to the fundamental and high reactance to all the harmonics. It follows that the output voltage V out is a sinusoidal waveform with a frequency equal to the switching frequency of the transistors. Due to the low power loss in the transistors, the class D power amplifier in general offers very high efficiency (>90%), which depends on the switching characteristics of the transistors. If the switching time of the transistors is very small, then the class D amplifier can provide good performances at the frequency range of the inductive link. Due to these interesting characteristics, researchers have proposed many approaches (12,13,16) based on the class D amplifier to power various biomedical prosthetics. The Class E Power Amplifier. The class E amplifier uses only one power transistor followed by a series resonant circuit (the primary circuit of the inductive link) (Fig. 10). The power transistor acts, in this case, as a low-loss switch. This will keep the amplifier’s power loss at a very low level. The primary circuit of the coupled coils and the capacitor Cl are tuned to a frequency close to the switching frequency of the transistor. This will result in a sinusoidal waveform at the output. The power efficiency of the class E amplifier can theoretically reach 100%. The class E power amplifier is used in many designs (5,25,26) to transfer power to prosthetic implants.


13

Main Inductively Coupled Link Approaches The desired high performances of transcutaneously powered devices are subject to parameter variations such as inductively coupled link misalignment, distance variations between emitter and receiver coils, operating frequency drift, and the change in electronic component values. But ideally, link performances such as the voltage gain and the power efficiency have to be insensitive to any of these changes. A great deal of interesting theoretical and experimental research has already been done in the field of the design of inductively coupled links dedicated to transferring power and data to prosthetic implants (12,13,16,18,19,21,26,27,28). Few of these available designs dealt with techniques for reducing the effect of variations of link parameters on its performances, however. Usually, these techniques focused primarily on the optimization of transcutaneous link performances, and secondly on minimizing the effect of link characteristic variations. The following sections describe briefly the main existing approaches that deal with inductively coupled link effect variations. The Geometric Approach. Galbreith, Soma, and White (12) proposed a common approach, called the geometric approach, which is used to reduce the effects of coupling variations. In this approach, the receiver and transmitter coils are of different sizes, with the receiver coil being the smaller one. In this way, the coupling factor is insensitive to the coil displacement, assuming that the receiver coil remains within the perimeter of the transmitter coil. The coils are allowed to move laterally, and even with small angular rotations, without significant changes in k. This interesting characteristic makes this approach tolerant to the various misalignments, but still sensitive to coil separation changes. The authors of this approach explain that for small separations, the coupling coefficient drops in proportion to the coil separation (12). And, since the voltage gain of the link is dependent on the coupling coefficient, this approach is not tolerant to the coil separation. Another problem with this approach arises from the fact that the transmitter and receiver coils are not of equal size. In this case, the coupling magnitude is smaller than that resulting from coils of equal size. With this reduction in the coupling magnitude, the current magnitude of the transmitter coil has to be high to obtain the same voltage gain as in the case of equally sized coils. But increasing the transmitter’s current leads to significant losses in the primary loop, which has to be avoided. The Stagger-Tuning Approach. Another approach that desensitizes the voltage gain of the inductive link to the coupling was also proposed in Ref. 12. This approach offers good displacement tolerance, efficiency, and large bandwidth. In this approach, the inductively coupled link is based on the stagger-tuning principle where both the receiver and the transmitter resonant frequencies are not equal and they are slightly different from the link operating frequency. The poles associated with these two frequencies are selected so that one pole is placed above the pole associated with the operating frequency and the other pole below. The position of these poles, however, is subject to the coupling changes. The key to this stagger-tuning approach is to place these poles in such a way that their movement will compensate for the link gain resulting from the coupling changes. For example, an increase in the coupling will normally result in an increase in the link gain, but at the same time the poles will move away from the pole associated with the operating frequency, which normally decreases the gain. These opposite effects cancel each other out and force the link gain to be insensitive to the coupling variations caused by the coil displacement. Therefore, the stagger-tuning approach is not only tolerant to the angular and lateral misalignment, but also to coil separation. Also, this pole placement increases the link bandwidth without affecting the link voltage gain and power efficiency. Another Approach. Another interesting design approach which is insensitive to coupling variations was proposed by Zierhofer and Hochmair (29). This approach focuses on optimization of the link power efficiency and coil misalignment tolerance. In this approach, a high-efficiency self-oscillating class E power amplifier is used to drive the transmitter-tuned circuit. The key to this design is that the amplifier oscillating frequency is not fixed, but affected by the coil coupling variations. The effect of coupling variations due to coil distance changes results in the changing of the reflected impedance of the receiver in the transmitter-tuned circuit. The variation in the reflected impedance is then used to control the oscillating power amplifier frequency. Zierhofer

14


Fig. 11. The inductive coupling stage circuit. The main blocks of this circuit are a class D power amplifier driven by an oscillator, a series inductively coupled link, and a voltage regulator.

and Hochmair have shown that the amplifier oscillating frequency tracks the absolute transmission efficiency maximum, thereby improving the power transmission against the coupling variation. Since the coupling is dependent on coil separation, this approach is tolerant to the coil distance variations. An Example of Applications. Figure 11 depicts an example of the schematic of a complete transmission energy and data circuit. It encloses a 20 MHz AM modulator that produces a carrier, the amplitude of which is modulated by the incoming Manchester-coded data (8). The resulting modulated carrier is then presented to a class D amplifier, the output of which is controlled by a power regulator. This device is intended to compensate for the variations in the coupling factor affected by the displacements between the transmitting and receiving coils. It increases the power voltage of the class D amplifier in the case of a high coupling and it reduces the power voltage in the reverse case (13). The secondary output signal of the transformer is then presented to a voltage regulator and an AM demodulator to provide coded data and power to the whole implant. By using such a system, two main objectives are sought: a minimum transmission bandwidth of about 1 MHz, since it has been designed for a maximum transmission rate of 500 kbit/s, and minimal losses to ensure the external power source durability of the future miniaturized system. To reach these two objectives, we have opted for a stagger-tuned link (12). Together with the power regulator technique, the link features low-energy consumption and possesses the advantage of a wide bandwidth (13) in addition to its adaptability to coil displacement. It also produces a fairly stable output voltage from which to recuperate a regulated voltage supply. The voltage regulator used in our system operates with a small voltage drop across it, so that it consumes little power while sourcing large currents, thus respecting the second consideration mentioned above. In the same vein, we note that the choice of the 20 MHz carrier depends on transformer efficiency, since a computer simulation model and experimental tests (Fig. 12) have indicated a maximum efficiency (24%) at this frequency.


15

Fig. 12. Variation of the power efficiency with the coil separation of the circuit presented in Fig. 11. The efficiency maximum is equal to 24% and is obtained at a frequency equal to 20 MHz and at the critical coupling of the inductively coupled link.

Summary and Future Developments With the recent developments in the microelectronics field, new complex and powerful implants with small sizes can be now achieved with minimum cost and in a short time. Unfortunately, we still have to overcome the critical problem of powering these implants. Therefore, there is an immediate need to develop new, smallsized powering systems with high efficiency. Since these implants are, in general, intended for long-term implantation periods, the use of batteries is to be avoided whenever it is possible. By avoiding the use of batteries to power implants, the surgery needed to replace the implant batteries becomes unnecessary, which will in turn reduce the risk of skin infection. From the present article, it is obvious that a powering system based upon an inductively coupled link presents a good choise for such applications. With this choice, the implant lifetime is no longer limited by the powering system lifetime, but only by the implant electronics. Since, with the use of an inductively coupled link, the energy transfer is achieved electromagnetically, the skin damage risk is kept to a very low level. The other interesting characteristic of the inductively coupled link is that it can be used to accomplish the dual tasks of bidirectional communication and energy transfer simultaneously. Currently, many researchers are developing new implants and new powering systems based on the class D and E power amplifiers and using inductively coupled links. The challenge is to integrate the implant electronics, the receiving power system, and the communication system in the same chip. In some cases, new storage elements are also to be included within the same chip to make the implant autonomous. The amount of energy transferred should be high enough in order to power the electronics of the whole implant. Finally, the powering system should a high-efficiency one so that the whole system remains economical.

16


BIBLIOGRAPHY 1. W. F. Agnew D. B. McCreery Neural Prostheses, Fundamental Studies, Englewood Cliffs, NJ: Prentice Hall, 1990. 2. W. Greatbatch C. F. Holmes History of implantable devices, IEEE Eng. Med. Biol. Mag., 10 (3): 38–41, 1991. 3. L. J. Seligman Physiological simulators: From electric fish to programmable implants, IEEE Trans. Biomed. Eng., 29: 270–284, 1982. 4. M. C. Shults et al. A telemetry-instrumentation system for monitoring multiple subcutaneously implanted glucose sensors, IEEE Trans. Biomed. Eng., 41: 937–942, 1994. 5. D. F. Williams Implantable prostheses, Phys. Med. Biol., 25: 611–636, 1980. 6. E. Feigenbaum Cochlear implant devices for the profoundly hearing impaired, IEEE Eng. Med. Biol. Mag., 6 (2): 10–21, 1987. 7. J. T. Mortimer Electrical excitability: The basis for applied neural control, IEEE Eng. Med. Biol. Mag., 2 (2): 12–13, 1983. 8. M. Sawan et al. Stimulator design, and subsequent stimulation parameter optimization for controlling micturition and reduction urethral resistance, IEEE Trans. Rehabil. Eng., 4: 39–46, 1996. 9. B. Smith P. H. Peckham M. W. Keith An externally powered multichannel implantable stimulator for versatile control of paralyzed muscle, IEEE Trans. Biomed. Eng., 34: 499–507, 1987. 10. R. B. Stein P. H. Peckham D. Popovic, (eds.) Neural Prostheses: Replacing Motor Function after Disease or Disability, Oxford: Oxford University Press, 1992. 11. M. Thoma The Leg Pacemaker, Vienna: MedImplant, Vienna, 1993, p. 4. 12. D. C. Galbraith M. Soma R. L. White A wide-band efficient inductive transdermal power and data link with coupling insensitive gain, IEEE Trans. Biomed. Eng., BME-34: 265–275, 1987. 13. M. Sawan et al. A new transcutaneous fully-programmable neural stimulator, Int. J. Microcomput. Appl., 13: 142–147, 1994. 14. D. C. Jeutter A transcutaneous implanted battery recharging and biotelemeter power switching system, IEEE Trans. Biomed. Eng., BME-29: 314–321 1982. 15. T. Tamura et al. Transcutaneous optical power converter for implantable devices, Proc. SPIE, 2084: 99–104, 1996. 16. N. De N. Donaldson T. A. Perkins Analysis of resonant coupled coils in design of radio frequency transcutaneous links, Med. Biol. Eng. Comput., 21: 612–627, 1983. 17. F. E. Terman Radio Engineering Handbook, New York: McGraw-Hill, 1943. 18. W. H. Ko S. P. Liang C. D. F. Fung Design of radio-frequency powered coils for implant instruments, Med. Biol. Eng. Comput., 15: 634–640, 1977. 19. Z. Tang et al. Data transmission from an implantable biotelemeter by load-shift keying using circuit configuration modulator, IEEE Trans. Biomed. Eng., 42: 524–528, 1995. 20. C. T. A. Johnk Engineering Electromagnetic Fields and Waves, New York: John Wiley & Sons, 1988. 21. E. S. Hochmair System optimization for improved accuracy in transcutaneous signal and power transmission, IEEE Trans. Biomed. Eng., BME-31: 177–186, 1984. 22. M. Soma D. C. Galbraith R. L. White Radio-frequency coils in implantable devices: Misalignment analysis and design procedure, IEEE Trans. Biomed. Eng., BME-34: 276–282, 1987. 23. H. Matsuki T. Matsuzaki A. Suzuki Energy transfer system utilizing amorphous wires for implantable medical devices, IEEE Trans. Magn., 31: 1276–1282, 1995. 24. C. M. Zierhofer E. S. Hochmair High-efficiency coupling-insensitive transcutaneous power and data transmission via an inductive link, IEEE Trans. Biomed. Eng., 37: 716–722, 1990. 25. Z. Hamici R. Itti J. Champier A high-efficiency power and data transmission system for biomedical implanted electronic devices, Meas. Sci. Technol., 7: 192–201, 1996. 26. P. R. Troyk M. A. K. Schwan Closed-loop class E transcutaneous power and data link for microimplants, IEEE Trans. Biomed. Eng., 39: 589–599, 1992. 27. I. C. Forster Preliminary development of a radiotelemetry system for biological applications, Med. Biol. Eng. Comput., 24 (3): 281–291, 1986.


17

28. T. Mussivand et al. A Transcutaneous Energy and Information Transfer System for Implanted Medical Devices, 1996, ASAIO Journal, 41 (3): M253–M258. 29. C. M. Zierhofer E. S. Hochmair Geometric approach for coupling enhancement of magnetically coupled coils, IEEE Trans. Biomed. Eng., 43: 708–714, 1996.

ABDELOUAHAB DJEMOUAI Ecole Polytechnique de Montréal MOHAMAD SAWAN Ecole Polytechnique de Montréal

51.Rehabilitation Engineering

Clinical Engineering Handbook (Biomedical Engineering)

Modern Control Engineering (Control Engineering)

Cell Engineering: Apoptosis (Cell Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Engineering Maintenance Management (Industrial Engineering)

Modern Ceramic Engineering (Materials Engineering)

Earthquake Engineering: From Engineering Seismology to Performance-Based Engineering

Earthquake engineering: from engineering seismology to performance-based engineering

Engineering Knowledge

Reckless Engineering

Engineering Infinity

Reckless Engineering

Tissue Engineering

Cryptographic engineering

Engineering Measurements

Engineering Vibrations

Neural Engineering

Engineering Ethics

Engineering Fundamentals

Engineering Knowledge

Engineering Infinity

Engineering Knowledge

Engineering Knowledge

Engineering Thermodynamics

Engineering Analysis

Engineering Mechanics.Statics

Biochemical engineering

Irrigation Engineering

Engineering mathematics

Electronic Engineering

51.Rehabilitation Engineering

Clinical Engineering Handbook (Biomedical Engineering)

Modern Control Engineering (Control Engineering)

Cell Engineering: Apoptosis (Cell Engineering)

Clinical Engineering Handbook (Biomedical Engineering)

Engineering Maintenance Management (Industrial Engineering)

Modern Ceramic Engineering (Materials Engineering)

Earthquake Engineering: From Engineering Seismology to Performance-Based Engineering

Earthquake engineering: from engineering seismology to performance-based engineering

Engineering Knowledge

Reckless Engineering

Engineering Infinity

Reckless Engineering

Tissue Engineering

Cryptographic engineering

Engineering Measurements

Engineering Vibrations

Neural Engineering

Engineering Ethics

Engineering Fundamentals

Engineering Knowledge

Engineering Infinity

Engineering Knowledge

Engineering Knowledge

Engineering Thermodynamics

Engineering Analysis

Engineering Mechanics.Statics

Biochemical engineering

Irrigation Engineering

Engineering mathematics

Electronic Engineering

Recommend Documents