A RECONFIGURABLE SYSTEM FOR MOVEMENT

REHABILITATION AND DIAGNOSTICS WITH FES

Piotr Kaczmarek, Andrzej Kasi

nski, Marek Kraft and Przemysław Mazurkiewicz

Pozna´n Univeristy of Technology, Insitute of Control and Information Engeenering, Piotrowo 3a, 60-395 Pozna´n, Poland

Keywords:

Electrical stimulation, rehabilitation, movement restoration, instrumentation.

Abstract:

The architecture of a custom originally designed FES system is described. The system is built from off-the-

shelf and cusom components to obtain a target functionality. List of potential applications is provided. Some

tasks have been already tested in laboratory and clinical conditions. Hardware speciﬁcation of proprietary

components is given and interfacing issues are addressed.

1 INTRODUCTION

Motion restoration through the appropriate functional

electrostimulation (FES) of a neuro-muscular system

is an interesting perspective. Still, the investigations

are continued at the level of a single joint as well as

that of a limb. The most interesting problem is to eval-

uate how the electrical stimulation program translates

into an efﬁcient and controlled elementary motion.

Commercially available functional stimulators can be

divided in 2 groups: stimulators for clinical research

(like the S88 from Grass Teledyne), daily-use electri-

cal stimulators for rehabilitation and functional elec-

trical stimulation systems used in neurophysiological

research (like NeuroTrac, Compex Motion, ODFS,

Parastep)(Keller et al., 2002; Taylor et al., 1999). In

principle all of these stimulators provide the possibil-

ity to alter the stimulation pulse frequency, amplitude

and duration. In most cases (S88, NeuroTrac, ODFS,

Parastep) these parametersare controlled manually by

the user. Only the most advanced stimulators, like

Compex Motion, allow to control the parameters by

means of a closed loop control based on data coming

from a number of sensors (switches, force sensors,

EMG sensors). Most common approach is a sim-

ple event-triggered stimulation for a certain amount

of time with the parameters set by the therapist (ie.

in ODFS stimulators the stimulation sequence is trig-

gered by a heel-switch). All of the above mentioned

stimulators lack the capability of optimizing the shape

of the single stimulation pulse, leaving only various

square or trapezoid pulse combinations at the thera-

pist disposal.

Many research institutes try to determine the in-

ﬂuence of stimulation parameters on the muscle con-

traction, the fatigue effect, as well as the long-term ef-

fects generated in the muscles subjected to persistent

stimulation (this refers to the regeneration process as

well as to side effects). The experiments are aimed at

investigation of physiological properties of the mus-

cles both of healthy and of disabled subjects (Chizeck

et al., 1999). An interesting issue is to reveal the opti-

mal stimulation pattern for the application in the FES

systems (Breen et al., 2006). So far the optimization

process was restricted to the selection of the stimuli

train frequency (Chou et al., 2005), or the pulse width

and amplitude. The inﬂuence of the stimulating pulse

shape on the contraction force has not been studied

yet. Moreover, research results of the experiments

concerning the pulse shape inﬂuence on the muscle

contraction are not clear (Bennie et al., 2002).

Both movement restoration and rehabilitation pro-

cesses are evaluated upon apparent effects. Such ef-

fects can be used as a feedback for the FES control

algorithm, however it is desirable to quantify the ef-

fects, which are of different character. First apparent

effect is the response of neuro-muscularsystem under

stimulation on its own. The neuro-muscular system

excitation level can be judged on the basis of the elec-

tromyographic signals (EMG) being the response to

Kaczmarek P., Kasi

nski A., Kraft M. and Mazurkiewicz P. (2008).

A RECONFIGURABLE SYSTEM FOR MOVEMENT REHABILITATION AND DIAGNOSTICS WITH FES.

In Proceedings of the First International Conference on Biomedical Electronics and Devices, pages 17-22

DOI: 10.5220/0001049500170022

 SciTePress

the electrostimulation. In movement restoration ex-

periments however, it is necessary to measure and to

analyze the EMG of the particular muscles contribut-

ing to the movement (Kutch and Buchanan, 2001).

Moreover, it might be desirable to use a multi-channel

transcutaneus electrode to support the decomposition

of the EMG signal for studying the propagation of

a single action potentials and motor point localiza-

tion, or to analyze the signals in more detailed way.

In order to get a better image of the stimulation ef-

fects there might be also required to partially esti-

mate the state of a biomechanical system from ad-

ditional physiological signals such as electroneuro-

grams (ENGs)(Sinkjaer et al., 2003) or mechanomyo-

grams (MMGs) (Kaczmarek et al., 2005; Orizio et al.,

2003) which better reﬂect the senso-motor system ac-

tivity as well as changes of the muscle geometry, re-

spectively.

However, these signals reﬂect the intermediate ef-

fects of electrostimulation, but the target result (the

ﬁnal effect) is the motion of a limb or at a lower level,

the motion of a joint under study. There are two in-

teresting aspects of the motion evaluation - the static

aspect, related to the force generated by a joint at a

certain perceived level of the EMG activity while cor-

related with joint conﬁguration (angle), and the sec-

ond, which is related to the resulting motion dynam-

ics. The investigation of the ﬁrst aspect is performed

by the simultaneous recording of EMGs of the appro-

priate muscle and the resulting force being directly

recorded with the tensometric device, while the joint

is locked at a particular angle. For the second aspect,

it is necessary to acquire simultaneously the signals

of the generated force, EMGs, the joint-angle rotation

signal and the task-space acceleration signal due to

the resulting dynamic motion of the limb.

This would be possible under the condition that

one disposes of a fully programable electrostimulator

with a necessary number of outputs, of a measurement

system running during stimulation (a multichannel

EMG signal acquisition and a processing system) and

of a measurement system to evaluate the biomechan-

ical effects of the FES (an instrumented exoskeletal

device - the orthosis). In this paper we describe such

a custom system, which has been built from modules

and list examples of applications of that system to the

rehabilitation and diagnostics procedures.

2 THE SYSTEM ARCHITECTURE

The schematic outline of the complete system is

shown in ﬁg. 1. The main control and sensors readout

tasks are performed by the embedded 3,5” PC. The

Figure 1: The block diagram of the FES system.

PC is equipped with a Intel Celeron 400MHz pro-

cessor, 256MB RAM and 4GB CompactFlash card.

The conﬁguration of the system involved in a partic-

ular task is open in a sense, that the use of a sim-

ple serial peripheral bus (SPB) and of a specially

designed communication protocol enables the user

to connect a number of intelligent sensors (like ac-

celerometers, goniometers, tensometers, point EMG

sensors) to form a distributed measurement system,

enabling tailoring of the system conﬁguration to the

particular task at hand. These distributed sensors

perform all the measurement, preprocessing and data

forming operations on-site. This allows to reduce the

inﬂuence of various environmental interferences on

measurements. The availability of sensor data enables

the system to perform real-time closed loop control

of joints by proper muscles stimulation depending on

this data. The embedded PC is equipped with a multi-

channel A/D converter PCM-3718HG. This converter

is used to measure the data from a multi-section EMG

electrode. The PC is also ﬁtted with the Bluetooth

wireless interface. This enables remote control (for

example by exchanging stimulation programs during

rehabilitation), the visualization of a system state and

sensor readouts (usable in biofeedback applications).

For simple therapeutic programs at the patient bed it

is possible to control the system by a palmtop.

2.1 Electrostimulator

The stimulator unit has the ability to perform stimula-

tion through four independent channels. This enables

independent stimulation of different muscle groups.

The control circuit consists of a microcontroller and

a 4-channel, 8-bit digital to analog converter (DAC).

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

Such the system setup enables the user to deﬁne the

pulse shape, frequency, amplitude and other parame-

ters (i.e. number of stimuli, initial delay, pulse train

proﬁle and modulation frequency) independently for

each stimulation channel. The output stage consists of

a voltage controlled current source, which can source

a current of up to 60 mA and can work with volt-

ages up to 400 Vpp, which is important in the case

of the transcutaneous stimulation. The high voltage

stage is galvanically separated from the control cir-

cuit. Control and programming of the device can be

performed on-line via RS232, SPI or I2C interface.

In that way the described device let the therapist to

deﬁne ﬂexibly the movement-restoration or rehabili-

tation requirements to each individual patient and to

the particular task.

2.2 Orthosis and its Instrumentation

The orthosis is a skeleton enabling the force and po-

sition sensors attachment. Fig. 2 presents the ex-

oskeleton of the ankle joint, which is one of the most

important joints for balancing and locomotion tasks.

The orthosis enables to perform static measurements

(during an isometric contractions), as well as dynam-

ical acquisitions during walking or balancing. It is

equipped with the angle sensor i.e. an incremental

encoder of 0.05 deg resolution and with the force sen-

sor based on a planar-beam force sensor with a full-

bridge strain gage. The system is capable of measur-

ing a torque generated by a dorsi as well as a plantar

ﬂexion during isometric contractions for variable an-

kle joint angles in the range between -10 and 10Nm.

The mechanical solution of the force sensor allows to

obtain the force, that is always perpendicular to the

sensor beam and independent on the ankle joint an-

gle. The orthosis controller has a 10-bit AD converter

with a variable input gain. This let the user to se-

lect appropriate gain depending on the performed test

and the torque value. Moreover, the controller per-

forms signal processing tasks such as a signal ﬁlter-

ing and sampling with rate up to 4kHz, evaluation of

the absolute angle of the joint and data buffering. The

communication and data transfer to the Embedded PC

is performed via SPB. Moreover, during locomotion

or balancing tests, 3D accelerometric sensors can be

ﬁxed to the exoskeleton. The accelerometric sensor

is equipped with a microcontroller and a 10bit AD

converter, moreover it has 3 digital inputs enabling to

connect the additional switches (foot switches or limit

switches) or other discrete triggering elements. The

smart sensor can operate in three modes, selected by

the user via SPB. The ﬁrst one is the data acquisition

mode, where raw data is transmitted to the embedded

PC. The second is the data processing mode, enabling

the estimation of the velocity and of the position and

the signal ﬁltering. The third is the events detection

or a fuzziﬁcation mode, which is used for fuzzy logic

control algorithm. The sensor in such a mode evalu-

ates the system state from the accelerometric signals

and digital input state according to the given set of

fuzzy rules.

The use of registered signals is the following. 3D

acceleration signal together with goniometric signal

let to reconstruct the position and the motion tra-

jectory independently. This enables to analyze the

kinematic effects of the electrostimulation and to dis-

cover pathological deﬁcits. The tensometric subsys-

tem gives the opportunity to evaluate the contraction

force at a single joint. Thus not only the kinestatic

evaluation is possible but also dynamics of the move-

ment can be studied.

Figure 2: Instrumented Orthosis of the ankle joint.

2.3 EMG Acquisition Subsystem

Multichannel EMG. Important part of the pre-

sented system is an EMG electrode (Fig 3). The elec-

trode consists of 9 rectangular AgCl contacts surfaces

placed of a distance of 5 mm and with the contact

surface size of 10x1[mm]. Multichannel EMG signal

acquisition system is modular, where every module is

connected to a single acquisition channel. It consists

of an instrumental ampliﬁer, analog high-pass Butter-

worth ﬁlter (f3dB=10Hz), low-pass Butterworth ﬁlter

(f3dB=1000Hz) and a ﬁnal ampliﬁer stage giving the

total gain within the range 1000÷2000. Acquired sig-

nals are collected by a PCM-3718HG card connected

to the embedded PC.

Point-like EMG Smart Sensor. Smart EMG sen-

sor is a device with analogue and digital parts. The

analog part is similar to that described in a previous

section. However, the contact surfaces (electrodes)

are integrated with the PCB of the analog subsystem.

A RECONFIGURABLE SYSTEM FOR MOVEMENT REHABILITATION AND DIAGNOSTICS WITH FES

Figure 3: Linear electrode array for multichannel EMG.

The digital part consists of microcontroller with a 10-

bits AD converter. Its main role is the acquisition of

the analog EMG signal, buffering and basic prepro-

cessing such as evaluation of the RMS value. The

sensor can operate in two modes. In the ﬁrst, the raw

data are transmitted via SPB to the master controller,

in the second mode only the results of processing are

transmitted. The sensor is designed to be placed di-

rectly over a tested muscle in order to reduce noise

and artefacts. Additionally, all IC chips can work in

the shutdown mode for protecting the sensor ampli-

ﬁers from damage due to the high-voltagestimulation.

3 APPLICATIONS OF THE

SYSTEM

3.1 Research Tasks

As far the system has been used in 4 research tasks

focused on the FES issues.

Stimulus Waveform Optimization. The system re-

stricted to the stimulator, the force sensor and the

EMG sensors has been used to evaluate the inﬂuence

of the stimuli shape on the muscle contraction dynam-

ics, and comfort level. The experiments were per-

formed on a group of healthy volunteers and subjects

with movement deﬁcits. Within this study the bipha-

sic stimuli with inter-phase interval (IPI) have been

tested. 24 stimulus waveforms have been taken into

account. They are combination of a pulse widths (50,

100, 175, 250us) and the inter phase intervals (0, 50,

250, 500, 1000, 2000us). The preliminary results sug-

gest that the contraction force recorded during stimu-

lation depends not only on the pulse amplitude, width

and frequency, but also on the IPI. The contraction

force has increased signiﬁcantly with increase of the

IPI. Additionally decrease of the pain sensation side-

effects was reported. The results suggest, that the

modiﬁcation of the pulse shape seems to be an alter-

native for the the commonly used force control tech-

niques such as a pulse-width, amplitude or frequency

modulation.

Evaluation of the EMG-force Relationship. Us-

ing the instrumented orthosis and exoskeletons one

is limited to apply the FES system only in the clin-

ics. Therefore it is essential to built the portable FES

with feedback based on measured physiological sig-

nals. This is necessary in order to evaluate or estimate

the force and the joint conﬁguration. In this study,

the maximal voluntary contraction force and the EMG

signal have been recorded while varying the ankle

joint angle. It was observed, that for a medial gas-

trocnemius muscle the maximal force increases with

increase of the dorsi ﬂexion, however the RMS value

of the recorded EMG signal signiﬁcantly decreases,

which may suggest the decrease of the muscle activ-

ity level. This phenomenon demonstrates that the for

the force estimation at a gastrocnemius muscle from

the EMG signal, the simultaneous estimation of the

ankle joint orientation is required. This effect did not

occurred for tibialis anterior muscle.

Evaluation of the Fuzzy Rules for a Fuzzy Logic

Controller for Movement and Balancing Tasks.

The efforts are made to create a fuzzy sets describ-

ing the phase of gate cycle during locomotion from

5 accelerometric sensors located at lower limbs and

the trunk as well as from EMG sensors. This could

enable to estimate the mass center position changes

during locomotion or balancing tasks. The smart sen-

sors are powerful enough to perform simple classiﬁ-

cation tasks and enable easy testing and evaluation of

the rules.

Developing of the FES Control Algorithms. Us-

ing the Embedded PC with Linux on-board as the

system supervisor enables an easy development and

testing of various control algorithms based on fuzzy

logic, neural networks or classical control. Moreover,

the standard, well known and powerfull tools and li-

braries can be used for the application developing and

testing. The distributed intelligent sensors may be re-

conﬁgured to work in a given control mode delivering

a appropriate feedback signals.

3.2 Diagnostics Tasks

The diagnostics tasks refer to the potential application

of the system in the clinical ﬁeld. In the presented

system it is possible to perform the electrophysiolog-

ical as well as kinematics test, which may support the

diagnosis. Moreover, such system could be used for

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices

Figure 4: The multichannel EMG signals recorded from a tibialis anterior muscle with 8-point linear electrode array. A

selected Action Potential (AP) propagating through the muscle has been marked with the arrows. The mean EMG signal

power has also been computed. The black dots denote the mean signal power of the particular EMG channels. The maximum

of the mean signal power between the channels ch4 and ch5 denotes the motor point localization.

selecting the individually optimized stimulation treat-

ment.

Multisection EMG. Signals recorded with the mul-

tichannel EMG signal acquisition system have been

used to estimate the conduction velocity of motor

unit’s action potentials (AP). This can be achieved by

a special technique based on analysis of the APs hav-

ing similar shape and being recorded in neighbouring

channels. Selected AP generated by a single MU has

been marked in Fig. 4. The AP propagation veloc-

ity can be computed on the basis of measured delay

of the AP observed in particular channels and known

distance between electrode points. Additionally, the

system allows to localize neuromuscular junction in a

particular muscle. It has been shown that signal pa-

rameters such as AVR or mean frequency (Fmnf) of

the signals recorded at given points located along the

tested muscle, can indicate the localization of the in-

nervation zone. The motor point can be localized in

two ways: the ﬁrst is localization based upon the max-

imum signal power and the second is the AP phase

inversion. The phase inversion can be seen in Fig 4

where the AP marked with the arrows is inverted in

the channels ch5, ch6 and ch7. The motor point lo-

calization is important for an appropriate stimulation

electrodes placement in FES application as well as a

linear array-electrode placement in the signal decom-

position task. The main purpose of the decomposition

task is to estimate a discharge time of a particular mo-

tor unit during a voluntary contraction. In contrary to

the decomposition systems of the intracellular EMG

signals, decomposition of the multichannel surface

EMG signals does not require to use needle electrodes

which is an advantage. Multichannel recordings may

compensate for the lower selectivity and give a deeper

insight into the motor units activity. These informa-

tion can be used in more sophisticated algorithms to

improve the overall performance. The smart EMG

sensors can be used during walking tasks in order to

A RECONFIGURABLE SYSTEM FOR MOVEMENT REHABILITATION AND DIAGNOSTICS WITH FES

analyze the muscles synchronization issues. Two sen-

sors located over 2 antagonistic muscles can discover

their synchronization during contractions. Moreover,

the information from these sensors can be used in con-

trol algorithms.

System Identiﬁcation. The estimation of the sys-

tem state only from physiological sensors or the de-

termination of fuzzy sets must be adjusted to the in-

dividual subject. Moreover, the optimal control algo-

rithm for the FES controller must be selected. The

system is capable to perform tests enabling to work

out the force-EMG relationship for each subject. It is

possible to prepare a set of tests which enable semi-

automatic calibration of the sensors system.

The Stimuli Optimization. The stimulation exper-

iments revealed the variability of the excitability level

for the same stimulation procedure. The variability is

dependent on the pathology but also on the individ-

ual features of the subject. Therefore, the stimulation

waveforms and stimulus shapes should be selected in-

dividually. It was observed, that the stimulation let to

divide the subjects into few groups. The system let to

perform tests enabling classiﬁcation of the subject to

the particular group.

3.3 Rehabilitation Tasks

Repetitive Exercises with or without Stimulation

Support. The aim of such a treatment is to increase

the maximum contraction force, or to increase a range

of motion. A subject can observe the actual force

level, the EMG amplitude, or the joint angle on the

screen, and to try to follow the reference trajectory as

set by the therapist. Moreover it is possible to use the

stimulation in order to compensate partially for the

error between the reference and the actual trajectory.

Restoration of Movement Functions for the Phys-

ically Disabled Subject. The main aim of the pre-

sented system is to develop a daily-use FES system

for the restoration of movement functions with a min-

imal number of sensors. The system is potentially ca-

pable to operate in closed feedback loop mode with

sensors and stimulation sequences conﬁgured indi-

vidually by therapist on the basis of the identiﬁca-

tion tests results. However, to obtain satisfying results

with this application further investigations and devel-

opment of control algorithm is necessary.

ACKNOWLEDGEMENTS

The work was partially supported by the Polish

Ministry of Education and Science, project no.

1445/T11/2004/27.

REFERENCES

Bennie, S. D., Petrofsky, J. S., Nisperos, J., Tsurudome, M.,

and Laymon, M. (2002). Toward the optimal wave-

form for electrical stimulation of human muscle. Eur

J Appl Physiol, 88(1-2):13 – 19.

Breen, P. P., O’Keeffe, D. T., Conway, R., and Lyons, G. M.

(2006). A system for the delivery of programmable,

adaptive stimulation intensity envelopes for drop foot

correction applications. Med Eng Phys, 28(2):177 –

186.

Chizeck, H. J., Chang, S., Stein, R. B., Scheiner, A., and

Ferencz, D. C. (1999). Identiﬁcation of electrically

stimulated quadriceps muscles in paraplegic subjects.

IEEE Trans Biomed Eng, 46(1):51 – 61.

Chou, L.-W., Ding, J., Wexler, A. S., and Binder-Macleod,

S. A. (2005). Predicting optimal electrical stimulation

for repetitive human muscle activation. J Electromyo-

graphy Kinesiology, 15:300–309.

Kaczmarek, P., Celichowski, J., and Kasi

nski, A. (2005).

Experimentally veriﬁed model of mechanomyograms

recorded during single motor unit contractions. J

Electromyogr Kinesiol, 15(6):617 – 630.

Keller, T., Popovic, M., Pappas, I., and Müller, P. (2002).

Transcutaneous functional electrical stimulator "com-

pex motion". Artiﬁcial Organs, 26(3):219–223.

Kutch, J. J. and Buchanan, T. S. (2001). Human elbow joint

torque is linearly encoded in electromyographic sig-

nals from multiple muscles. Neurosci Lett, 311(2):97

– 100.

Orizio, C., Gobbo, M., Veicsteinas, A., Baratta, R. V., Zhou,

B. H., and Solomonow, M. (2003). Transients of the

force and surface mechanomyogram during cat gas-

trocnemius tetanic stimulation. Eur J Appl Physiol,

88(6):601–606.

Sinkjaer, T., Haugland, M., Inmann, A., Hansen, M., and

Nielsen, K. D. (2003). Biopotentials as command

and feedback signals in functional electrical stim-

ulation systems. Medical Engineering & Physics,

25(2003):45 – 96.

Taylor, P. N., Burridge, J. H., Wood, D. E., Norton, J.,

Dunkerly, A., Singleton, C, and Swain, I. D. (1999).

Clinical use of the odstock drop foot stimulator - its

effect on the speed and effort of walking. Arch Phys

Med Rehabil, 80:1577 – 1583.

BIODEVICES 2008 - International Conference on Biomedical Electronics and Devices