hammer in the shoe of the subject. .... It is possible to wear a pair of shoes with the carbon fiber ..... the active mode the two joints could slip, if the subject was.
EEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 3, NO. 4, DECEMBER 1995
299
An Actuator System for Investigating Electrophysiological and Biomechanical Features Around the Human Ankle Joint During Gait Jacob Buus Andersen and Thomas Sinkjaer, Member, ZEEE
Abstract-A system has been developed which is able to impose a fast ankle rotation during gait. The two-link system consists of a mechanical joint which is strapped to the calf and the foot of the subject. The mechanical joint turns around the anMe joint by means of bowden wires and it is connected to a motor placed next to a treadmill where the subject is walking. By position feedback from the mechanical joint the motor is regulated in such a way that it follows the movement of the ankle joint without influencing the gait pattern. The system is designed to impose a fast ankle rotation with a displacement of up to 20", e.g., a 8" stretch is performed with a rise time of 32 ms during any time of the gait cycle. The system attached to the subject's leg weights in total 0.9 kg. The system is designed to elicit a stretch reflex in the ankle extensor muscles during walking. The prospect of the system is to investigate electrophysiological and biomechanical features during a naturally evoked stretch reflex of the ankle extensors in both healthy and motor impaired subjects. The system is able to produce a perturbation of the ankle extensors at a walking speed of up to 6 km/h during the total gait cycle.
I. INTRODUCTION
E
XPERIMENTS with animals [l], [2] and with both healthy and motor diseased humans [3], [4] have shown the mechanical importance of the stretch reflex during sustained muscle contractions.It was found that the stretch reflex contributes to the total stiffness of the muscle with up to 50%. This implies that the stretch reflex has to be modulated during gait to achieve a sufficient muscle coordination [5]. To study the electrophysiology and biomechanics during a naturally evoked stretch reflex of the human ankle extensors, a rapid stretch has to be applied to the ankle joint during walking, e.g., by a device attached to the ankle joint. It is important that the device does not influence the natural pattern of gait, i.e., the weight and dimension of the system should be minimized. At any time during the gait cycle it should be possible to apply a well-defined stretch. Various methods have been described with the aim of inquiring into the role of the reflexes. Different mechanical systems have been developed to measure the static force in the human ankle joint under static conditions, such as sitting or standing. Some of the systems were designed for subjects sitting in a chair with their foot strapped to a platform
connected to a torque motor, which made it possible to rotate the ankle joint of the subject [6]-[8]. Nashner [9] described a hydraulic platform system with the ability to induce both a lateral and a rotational motion around the ankle joint of a subject standing. Several authors have attempted to investigate the stretch reflex for humans during gait. Yang et al. [lo] developed a method to apply a muscle stretch during walking. It consisted of a pneumatic system which was able to apply a stretch to the triceps surae muscle. However, the system was limited to the early stance phase before heel-off [ll]. Llewellyn et al. [12] investigated the transmission of the human tendon jerk reflexes during stance and gait by incorporating a spring loaded hammer in the shoe of the subject. The hammer was activated on the Achilles tendon. The set-up was not able to investigate the biomechanics of the ankle joint and to make reproducible stretches of the tendon. So far, none of the traditional systems have been able to deliver a reproducible perturbation, which could explore the stretch reflex and the true kinetics of the human ankle joint during the entire gait cycle. The aim of this work has been to develop a mechanical system which is able to deliver a well defined stretch throughout the entire gait cycle. The results will be used to investigate electrophysiologicaland biomechanical features of the human ankle joint during dynamical conditions and to investigate how much the stretch reflex is contributing to the control of locomotion.
11. EQUIPMENT
The developed system is constructed for a body weight supporting treadmill. An actuator system involving an acmotodgear system is placed beside the treadmill on an adjustable table (see Fig. 1). By means of bowden wires the torque of the actuator is transferred to a functional joint attached to the ankle of the subject. The functional joint is mounted on two carbon fiber casings strapped around the calf and the foot. The control system of the motor is able to operate in two modes; a passive and an active mode. In the passive Manuscript received July 6, 1995; rewsed August 29, 1995. This work was mode, the position of the motor is locked to the position of the functional joint in order to follow the movement of the ankle supported by CAMARC-I1 and The Danish National Research Foundation. The authors are with the Department of Medical Informatics and Image joint without interfering with the natural pattern of the gait. In Analysis, the Center for Sensory-Motor Interaction, Aalborg University, the active mode the system is able to perturb the ankle by a Aalborg DK-9220, Denmark. predefined rotation of the joint. IEEE Log Number 9415967. 1063-6528/95$04.00 0 1995 IEEE
IEEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 3, NO. 4, DECEMBER 1995
300
Actuator
T’r
A
-4
Cut A-A
U
Fig. 1. Treadmill system and adjustable actuator table. The mechanical joint attached to the ankle of the subject is turning around the ankle joint and is connected to the motor placed next to a treadmill by means of bowden wires. By a position feedback from the mechanical joint the motor is regulated in such a way that it follows the movement of the ankle joint without influencing the gait pattern. When required, the system is designed to impose an anlde rotation.
In order to achieve a dynamic range of measurement the system must be able to follow the ankle joint rotation of a subject walking with a fast cadence. The kinematics of the human ankle joint have been investigated during gait at 2, 4, and 6 km/h by the use of an optical goniometer. The maximum plantar flexion was -31Oo/s and the maximum dorsal flexion was -2OO0/s at a cadence of -6 km/h. At a cadence of -4 km/h the maximum acceleration was 15 OOOo/s2. According to the lunetic studies of Winter the maximal mean ankle joint moment of force is 1.74 “ k g f 0.22 std. (n = 17) [13]. During a dynamic movement of the ankle it can be sufficient just to lock the movement of the ankle to elicit a stretch reflex. If the maximal performance of the system is set to 200 Nm this will limit the weight of the subject to 115 kg (200 Nd1.74 “ / k g ) when a maximal counter action joint moment of force has to be performed. A force beyond this torque must be applied to gain an additional stretch of the muscles. In order to follow the ankle joint without resisting the movement, the system has to perform an ankle rotation of up to 310°/s and an acceleration ol’ up to 150OO0/s2 synchronously with the ankle joint.
Fig. 2. Functional joint constructed with two joints aligned on a fixed axle with six independent slide bearings. Three transducers are placed in this joint. A Wheatstone strain gauge bridge is measuring the resulting perturbation force in the beam attached to the casing of the foot, and two optical encoders with a resolution of 1000 CPR are mounted in the functional joint. The first encoder measures the actual position of the ankle joint and the other encoder the hysteresis angle. (1) Beam attached to the calf. (2) “Fork” attached to the foot. (3) Pulley attached to bowden wire. (4) Bowden cable. ( 5 ) Inner wire of bowden cable. (6) Optical encoders. (7) Strain gauge. (8) Joint acting on the ankle joint. (9) Joint connected to the bowden mechanism. (10) Hysteresis angle between the two joints.
The hysteresis angle is imposed to avoid any influence from the motor and the transmission in the passive phase. An individual plaster cast of the leg is made for each subject and a carbon fiber reinforced epoxy casing is casted to obtain a unique interface from the functional joint to the ankle of the subject. The carbon fiber casing is divided into two casings which are strapped with Velcro to the calf and the foot of the A. Functional Joint subject (see Fig. 4). The functional joint is mounted on the The task of the functional joint in the passive mode is to casings with the fixed axle aligned in level with the left ankle follow the movement of the ankle without interfering. The task joint (about a centimeter ahead of and a centimeter below the of the active mode is to transfer a rotation from the actuator lateral (fibular) malleolus [14]). to the ankle joint to impose a stretch of the ankle extensors The weight of the functionaljoint attached to the foot is 884 or flexors. The functional joint (see Fig. 2) is a construction g. It is possible to wear a pair of shoes with the carbon fiber with two joints which ’are aligned on a fixed axle with six supported joint. The right shoe is supported with an insole to slide bearings. The first joint consists of the “fork,” which is compensate for the extra height of the left foot caused by the mounted to the foot, and has one degree of freedom in relation carbon fiber. to the beam attached to the calf. The second joint consists of Three transducers are placed in the functional joint. A a pulley, which also has one degree of freedom in relation to Wheatstone strain gauge bridge (2 x HBM1.5/120XY21) is the beam attached to the calf. The second joint is connected to measuring the force from the perturbation in the beam attached the actuator by the bowden transmission. In addition, this joint to the casing of the foot. The strain gauge signal is amplified is coupled to the first joint with a hysteresis angle of k3.5’. by a precision strain gauge amplifier (SGA200).
301
ANDERSEN AND SINKJBR. SYSTEM FOR INVESTIGATING FEATURES AROUND THE HUMAN ANKLE JOINT DURING GAIT
Two optical encoders (HP-HEDS 9000 B, HEDS 6100) with a resolution of 1000 CPR are mounted in the functional joint. The first encoder measures the actual position of the ankle joint, and the other encoder measures the hysteresis angle. The resolution is increased to 4000 CPR (11.11 countsdeg.) by means of quadrature circuit.
the inner cable (ICi) and the outer cable (IC,) can be calculated, when knowing Young’s modulus of the material and the crosssectional area.
kN
k , = A , . E, = 11.56mm2 * 210 -= 2428kN. mm2
B. Bowden Mechanism The Bowden mechanism connects the functional joint to the actuator system. The Bowden cable is a bendable power transmission element which is only able to transmit tensile forces. It consists of two cables; an inner wire and an outer cable constructed as a steel spiral which provides compressional strength. The Bowden cable is a flexible, light, and powerful transmission. However, the friction of the cable causes a considerable loss of energy in the transmission. The dimensions of the inner cable rely on the actual tensile force, mechanical abrasion, bending fatigue, radial pressure, and corrosion. The strain and stress distributions in the wires are very complex wherefore the wires are dimensioned from the maximum load multiplied with a proper safety factor. A relatively low safety factor of 1.5 has been chosen (2.6 mm 7 x 7 right-regular lay) to reduce the bending stiffness of the cable. The friction coefficient of the cable is dependent on the curvature, the curve length, and the forces acting on each end of the cable. Assuming that the friction of a bowden cable follows the behavior of coulomb friction we have [15]:
F
/ K
ds
(g)
(1)
where 1) F0 is the force in one end of the cable = mog; 2) F is the force in the other end = mlg; 3) p is the friction coefficient; 4) IE is the curvature of the cable. The friction coefficient was found by hanging two equal weights at each end of the inner cable. A bag was attached to one of the weights and filled with sand until the cable starts moving. The force acting on this bag (F),the initial force (PO), and the curvature of the cable are used to find the friction coefficient. The force was varied from 28-1100 N in nine experiments . with a curvation IC, ds which was varied from T to 2 ~The friction coefficient was found to 0.127 f 0.012. The maximal bending of the bowden cable during gait is estimated to a curvature IE ds = T . With a friction coefficient of 0.13 the loss of transmitted energy in the bowden cables is up to 33% according to (1). When the system has to perform a perturbation at a high torque, the actuator has to compensate for elastic elongation of the bowden cables. To calculate the elongation of the bowden cable the stiffness of the bowden cable has to be known. The bowden cable can be considered as two springs in series. The specific stiffness (Nlrelative change of length) of
s
+
IC,
IC
IC0
C. Actuator System The actuator system is mounted on an adjustable table placed next to the treadmill. It is possible to adjust both the height and the angle of the motodgear system. The chosen motodgear system has been selected on the basis of the specified requirements to acceleration and velocity in the passive mode and to the necessary velocity and torque of the perturbation in the active mode with respect to the friction of the transmission. The chosen system is a brushless ac servo motor with a matched ac digital sinusoidal servo drive (95DSE92300 Dutymax ac-servo motor 2.6 kW and DB420 Digitax ac-servo amplifier 3 x 380 V IP20 8.5 N17A) and a planetary transmission (Harmonic Drive HPGP36 I = 12 : 1). This combination gives a rated torque of 100 Nm and a peak torque of 331 Nm of the output shaft of the gear. Due to the friction in the transmission, the maximal torque, which can be transferred to the functional joint in the active phase, is 218 Nm (331 Nm x 0.66) when the curvature of the bowden cable is K ds = T . The maximal acceleration, which the system in able to perform at the foot of a torque of 200 Nm, can be found by calculating the equivalent moment of inertia of the complete system
s
= F , , ~ U S*~ ~ S
1 p=-----ln
The specific stiffness of the bowden cable can be found as 1 1 1 - = - - @ IC = 511.5kN.
2 Wmotor
pa
(%)
Wmotor 2
Wmotor
+
Igeav
+ Imotor.
Where the moments of initia have been calculated to the following. 1) Foot 15.0. lop3 kgm2 [SI. 2) Pulley at foot 0.8 . lop3 kgm2. 3) Pulley at actuator 2.1 . kgm2. 4) Motor 0.6 . kgm2. 5) Gear 0.05 . kgm2. Which gives
Ieq= 7.7.lop4 kgm’. With respect to the friction of the bowden cable a torque of 200 Nm at the foot gives a required torque of 300 Nm at the shaft of the gear which equals a moment of 25 Nm with the chosen gear at the motor. The maximal acceleration of the motor at this load is: amator
=
M m o t o r , max - M m u s c l e Ieq
rad
= 3377 S2
302
IEEE TRANSACTIONS ON REHABILITATION ENGINEERING,VOL. 3, NO. 4, DECEMBER 1995
which gives a maximal acceleration at the ankle of a f o o t , max
=
amotor, max
12 rad = 281.4 S2
deg = 32246 S2.
The force acting on the bowden wire is 6624 N when performing a torque of 331.2 Nm through an arm which equals the radius of the pulley (T = 0.05 m). From this force and the known specific stiffness of the bowden wire the elongation of the cable can be found
F=k.&* F E=-
IC
-- 6.6kN
511.5kN
= 0.013 A1 &=-e 1 A12m = E .I = 0.013 . 1.6 m = 0.02 m.
(2) The found elongation corresponds to an angular movement of 23O, for which the controller has to compensate. The maximal elongation for a typical subject of 80 kg is 15". The stiffness of transmission can be expressed as 278.9 N/". D. Motor Control
The motor control consists of two controllers; an internal controller in the ac-servo drive and an external controller. The internal controller is coupled as a speed controller with feedback speed reference from a resolver in the servo motor. A reference from the external controller is algebraically summed with the speed feedback signal to produce an error signal. The error signal passes a PID filter to produce a current reference. The current reference is compared with the actual current in the motor and with the resolver feedback to produce a final current reference for a pulse modulated control of the power output stage (see Fig. 3). The external controller is a software implementation (language C) on a PC (Pentium 66 MHz IBM compatible PC) equipped with an I/O board (Data Translation DT2811). The controller is working in two modes; a passive and an active mode. In the passive mode a PID controller is implemented with a feedback signal from the optical encoder measuring the hysteresis angle in the functional joint. The aim of the controller is to keep the hysteresis angle in an equilibrium state. The controller begins each session with an initialization of the system to know the exact hysteresis angle, the maximal rotation, and the absolute end position. This initialization procedure is done by rotating the shaft of the gear pulley by the motor to the electric security switches. By means of a trigger signal from the data collection module the controller switches from the passive to the active mode. The angle, speed, and hold time of the perturbation in the active mode have to be specified before the experiment is
l _ _ _ - _ _ _ - - - - - - - - _ - - - - J
Fig. 3. Principle of intemal and extemal motor controller. The intemal controller is coupled as a speed controller with feedback speed reference from a resolver in the servo motor. The extemal controller is implemented as a PLD controller with a feedback signal from the optical encoder measuring the hysteresis angle with the goal of avoiding the two joints in the functional joint to interact.
started. The speed of the perturbation ranges from O-60O0/s. When the control module is triggered from the data collection module, it measures if the angle is too close to the end-stop to make a perturbation and, if not, it makes a rotation of the ankle with the predefined parameters and returns to the PID controller. The software security system limits the angle of the stretch to 20". The timing parameters of the perturbations are controlled by the data collection module. E. Dutu Collection The aim of the data collection module is 1) to collect electromyographical(EMG) data from the muscles, force and position from the functional joint, and timing parameters from the foot switch and 2) to trigger the motor control module with a proper timing for the perturbations of the gait. The total system is shown in Fig. 4. EMG's are recorded from the Soleus (SOL) muscle and Tibialis Anterior (TA) muscle with bipolar surface electrodes. The EMG's were amplified and filtered from 20-1000 Hz (DISA, model 15COl). Torque, angular position and EMG signals were fed to PC (486-66 MHz IBM compatible PC) equipped with a data collection module (Data Acquisition Card PCL718) at a sampling rate of 2 kHz.The EMG's were digitally rectified and filtered from 0-20 Hz. A foot switch is placed under the heal of the subject. In the active mode the system is able to make a rotation of the ankle joint with a timing triggered by the foot switch. The ankle is perturbed at random times in the gait cycle. During an experiment the perturbation is normally applied infrequently at a rate of one perturbation every 5-10 strides. Since the motor system has to operate simultaneously with noise-sensitive electrophysiological measurements, the electromagnetic noise of the motor system has been reduced by using a RFI filter and ferrite cores on the supply connections. Shielded wires and proper ground connection have been implemented throughout the entire electrical circuit.
ANDERSEN AND SINIUBR SYSTEM FOR INVESTIGATING FEATURES AROUND THE HUMAN ANKLE JOINT DURING GAIT
,
d-!==,
303
no. 3 did not show any significant changes when wearing the system. For all subjects the rectified, filtered, and average EMG Computer activity in the late stance phase measured over 80 ms was 4 not significantly decreased ( P > 0.28, Wilcoxon signed ranks test) when wearing the system. In the late swing phase the activity was initiated 2.6 f 1.6% earlier in relation to the complete gait cycle with the system attached than without the system attached. The TA onset in the late swing phase was generated significantly earlier with the system attached ( P < 0.05, Wilcoxon signed ranks test). An experiment was performed where a mass was added to the leg of a subject corresponding to the mass of the attached Amplifier system. The mass of the system could not explain the earlier onset of the TA EMG in the late swing phase, but when the subject was wearing the system it became clear that the tighter the Velcro was strapped to the leg, the more significant the change in EMG became. The velocity, which the system was able to perform during a perturbation, was generally lower during the stance phase than during the swing phase. Since the primary afferents from the muscle spindles are very sensitive to velocity [16] it is of severe importance that the stretch velocities are comparable throughout the gait cycle. The problem with maintaining comparable stretch velocities was solved by running the experiments with the control unit set at different stretch velocities and by matching the intervals in the step with regard to Fig. 4. Motor control and data collection.The EMG from the muscles, force velocity throughout the gait cycle. In the stance phase at a and position from the functional joint are fed to a computer unit. From the timing parameters of the footswitch the motor control module is triggered for walking speed of 6 km/h the maximal stretch performance of the perturbations of the gait. the system was 244"/s for an 8" stretch at a subject with a body weight of 85 kg. Three security systems are present to prevent the powerful During an experiment the gait cycle was divided into 10 system from evoking a stretch beyond the limits of the ankle; intervals, and a stretch of 8' with a velocity of 300"/s and a software, an electrical, and a mechanical stop. The electrical with a hold time of 200 ms was applied at random at a rate of and the mechanical security systems are mounted on the pulley one perturbation every 5-10 strides in each interval. A stretch at the shaft of the gear and are adjusted to the range of the during the stance phase and during the swing phase is shown ankle joint of each subject. in Fig. 6. Time 0 correspondsto 270 ms after heel contact (HC) where III. RESULTS a stretch is applied. After the induced stretch of the SOL a To investigate the effect of the attachment the EMG profiles reflex (marked with an arrow) occurs with a latency of 50 of the SOL and TA were investigated in six healthy subjects ms as expected for a short latency stretch reflex. No major (age 25-36, weight 50-85 kg). All subjects with the system interaction of the antagonist TA is seen as a consequence of attached showed a clear modulation of the SOL and TA EMG the applied stretch of the agonist in the stance phase. After the during a step cycle as described by Winter [13], and the EMG 200 ms holdtime the position of the ankle returns within 100 profiles with the system attached reflected intact patterns of ms to the same timing as without the perturbation but with an activity compared with the EMG profiles without the system offset of -2 degrees. After the initial force of the perturbation in the stance phase the force is decreasing as an effect of the attached. An individual casting was made for three of the subjects small voluntary dorsal flexion. In the swing phase where time 0 correspondsto 800 ms after to achieve a proper interface between the subject and the mechanical system. In Fig. 5 two EMG patterns of the SOL heel contact, a stretch reflex occurs with the same latency as in the stance phase. Coherent with the stretch reflex of the SOL and TA are shown when walking with a normal cadence (-3.5 km/h) with and without the perturbator attached to the leg. in the swing phase a depression is followed by a facilitation in Subject no. 6 is a 36-year-old healthy female subject who the TA. The position of the ankle returns within 200 ms after shows the most significant changes in the EMG. The activity of the release of the ankle to the exactly same level as without the SOL is decreased with 8% in the late stance phase whereas the perturbation. After the initial forces, the force is gradually the timing of the SOL EMG remains unaffected when wearing increased to a plateau after 180 ms probably because of the the system. The EMG activity of the TA is initiated 3.6% plantar extension during the stretch and because of the stretch earlier in the swing phase when wearing the system. Subject reflex. Motor Control
User Interface
Data collection
EEE TRANSACTIONS ON REHABILITATION ENGINEERING, VOL. 3, NO. 4, DECEMBER 1995
304
Without System
10
0
-10
!
-20 0
.
.
200
400
600
800
1000
T i e [ms]
-20;
200
400
600
800
1000
Time [ms]
Fig. 5. Full-wave rectified, filtered and average (n = 20) EMG activity of the Soleus and Tibialis Anterior muscles for subject no. 6 and subject no. 3 during walking with and without the system attached. The position of the ankle joint is shown with the attached system. A positive angle increment corresponds to a plantar flexion.
to transfer the forces of the motor in the active phase to the ankle joint of the subject. The friction of the bowden wire was found to result in a loss of transmitted energy of up to 33%. This limits the system to perform a torque at the ankle of 218 Nm with a maximal IV. DISCUSSION acceleration of the system at a torque of 200 Nm of 32000 A system able to apply a fast ankle rotation throughout the deg/s2. At this load the elastic elongation of the cable was gait cycle was developed. The system consists of a mechanical 0.02 m. Stiffer cables have been considered but these cables joint strapped to the leg of the subject. The mechanical joint is were both heavy and unhandy and did affect the walking in turning around the ankle joint, and by means of bowden wires the passive phase instead. Since the position feedback to the it is connected to a motor placed next to a treadmill where controller is frpm the functional joint this elongation does not the subject is walking. This design seems to be an effective give any position error at the ankle, but one has to compensate tool for evoking stretches of the human ankle extensors during for the error when setting the security switches at the actuator. The system was able to produce an 8" dorsal ankle rotation dynamical conditions. With the attained system it is possible of typically up to 50O"ls during the swing phase where the load to investigate the ankle joint mechanics and to examine how the stretch reflex contributes to the control of locomotion was typically low and up to 30O"ls during the stance phase where the load was high. At the highest load the maximal throughout the total gait cycle. Any attempt to measure a complicated physiological in- performance of the system was in the middle of the stance teraction, such as the peripheral motor control, changes the phase for the subject with the highest body weight at a walking conditions under which it is measured. The equipment de- speed of 6 kmk.At this point the system was able to perform scribed here is developed with special reference to the goal an 8" stretch with a maximal stretch velocity of 244"/s. All of minimizing the load of the system. The influence, which investigated intervals in a step were matched with regard to the the device added to the subject, had a minor effect on the highest possible stretch velocity at maximal load to maintain walking pattern. A significantly earlier onset time of 2.6% of comparable stretch velocities throughout the gait cycle. The the step-duration for the TA muscle was found by comparing highest possible dorsal stretch velocity was 120"ls higher than the walking pattern of the subjects before and after wearing the the voluntary induced dorsal ankle velocity. All kinematic parameters are measured in the functional functionaljoint. The duration of the step cycle was unchanged. The tight Velcro strapped around the leg was the major reason joint to avoid the error produced by the bowden transmission. for this phenomenon. The tight Velcro was needed in order However, the high torque generated by the ankle extensors As a result of the preactivated muscle and the initial length the initial peak in the force caused by stretching the passive tissue and the intrinsic properties of the muscles [8] in the stance phase is more than twice the size as in the swing phase.
ANDERSEN AND SINWER: SYSTEM FOR INVESTIGATING FEATURES AROUND THE HUMAN ANKLE JOINT DURING GAIT
80
1SOLEMG [PVI
'
I
Stance Phase .H
SOLEMG
80
Swing Phase '
305
-
-
With Perturbation
I
Without Pernubation
60
20
.
h
.^ 4u
-
0-
I
I
120
SOlPosition
1
;
:
l?z!zc
-1 0
Torque
: 10
0 -10'.
I
I
20
I 10
I
d
\ 400
42b0
0
200
400
600
Time [ms]
0.200
600
200
Time [ms]
Fig. 6. EMG activity of the Soleus and Tibialis Anterior muscles and position of the ankle joint shown with and without a perturbation during walking (5 W t ) . The resulting torque is shown for the perturbation. The examples are (a) during the stance phase (time 0 corresponds to 400 ms after HC) and (b) during the swing phase (time 0 corresponds to 800 ms after HC). Both perturhations are of go, 30O0/s, and with a hold time of 200 ms. With a latency of 52 ms after the perturbation a stretch reflex is evoked in Soleus muscle.
during walking [13] makes it likely that part of the stretch was absorbed by the tissue around the ankle joint. This means that the stretch of the ankle extensors was less than the stretch of the functional joint. However, it has been impossible to mount a goniometer on the ankle independently of the system to measure this error, but the casted bandage has been designed with the aim of minimizing this error. Visual inspection during the experiments and video recordings demonstrate that a perturbation of the functional joint was transmitted to the ankle joint at all the phases of the step. The functional joint was constructed as two joints. One joint was attached to the calf and the foot and acting synchronously with the ankle joint. The other joint was connected to the motor by means of bowden wires and also acting on the ankle joint but with a hysteresis angle of f3.5'. By controlling the actuator and thereby making this joint stay within the hysteresis angle, it was prevented that the motodgear system would disturb the natural pattern of gait in the passive mode. A control system based on a position feedback, where the ankle joint is directly coupled to the actuator, has to rely on the compliance of the transmission in such a way that it is possible to perform a change of state in the position encoder at the ankle joint and from this change accomplish an equivalent change in the position of the actuator. In addition to a significantly high compliance to avoid resistance from the motor, the system
requires a very fast and precise regulation without any kind of delay. All problems of the passive phase are eliminated by applying a hysteresis angle between the ankle joint and the mechanical system. By controlling the functional joint attached to the bowden wire to stay within the limits of the hysteresis angle of f3.5", the ankle joint can move without interference from the actuator system. A disadvantage of the hysteresis angle is that the angle has to be bridged in the active phase when making a perturbation. This could induce problems with delay and a collision in the joint. However, the delay is constant and the collision does not cause any artifact in the measurements probably because of the elasticity of the system. When doing a dorsal perturbation in the active mode the two joints could slip, if the subject was doing a voluntary fast dorsal flexion at the same time. This was reflected in the measurement of the force since the force transducer only measured the perturbing force when the two joints were coupled together. Yang et al. [lo] have described the only system known to the authors capable of investigating stretch reflexes during walking. The system consisted of a pneumatic device attached to the shoe of the subject. The system was weighing 1.5 kg and it could produce a stretch of between 3-9" with a peak velocity of 40-100"/s during the first 5040% of the stance phase. The system described in this paper weighs 0.9 kg and it
IEEE TRANSACTIONS ON REHABILITATION ENGINEEIUNG, VOL. 3, NO. 4, DECEMBER 1995
306
can produce a stretch of between 1-20’ with a peak velocity of 244-50O’ls during the complete gait cycle. A major difference in the mechanical characteristic of the perturbation done by the pneumatic system and the system described in this paper is that the pneumatic system tends to extend the knee joint when lifting the forefoot and thereby causing a postural sway of the body, whereas this perturbation system tends to flex the knee joint. The reason for this is that the bandage is both pulling the shin forward and the foot upward, when doing a dorsal perturbation, and is thereby not causing a postural sway. Future goals would be to investigate the peripheral motor control of the human ankle joint at dynamical conditions during normal and pathological motor behavior and to compare the results with those achieved by the artificially evoked Hreflex [17]-[19]. Since the risetime of the stretch is faster than the latency of the reflex, it is possible to investigate the intrinsic properties of the muscle during gait as well as how the stretch reflex modulation is contributing to the control of locomotion.
ACKNOWLEDGMENT The authors would like to thank Associate Prof. S. H. Nielsen, the Department of Production, Aalborg University, Assistant Prof. M. R. Hansen. Assistant Prof. 0. Jensen and Laboratory Tech. L. Jacobsen, Inst. of Mechanical Engineering, Aalborg University, for ideas and construction of the functional joint. A special acknowledgement to Orthotics and Prosthetics Jens Ditlev, Sahva, Aalborg, for ideas and construction of carbon fiber supports. The authors would also like to thank M.Sc.EE. K. Larsen, Inst. of Electronic Systems, Aalborg University, for the development of the data collection software and his quick service in times of panic. REERENCES [l] T. R. Nichols and J. C. Houk, “Improvement in linearity and regulation of stiffness that results from actions of stretch reflex,” J. Neurophysiol., vol. 29, pp. 119-142, 1976. [2] J. A. Hoffer and S. Andreassen, “Regulation of soleus muscle stiffness in premammillary cats: Intrinsic and reflex components,” J. Neurophysiol., vol. 45, pp. 267-285, 1981. [3] E. Toft, T. Sinkjar, and A. Rasmussen, “Stretch reflex variation in the relaxed and the pre-activated quadnceps muscle of normal humans,” Acta Neurol. Scand., vol. 84, pp. 311-315, 1991. [4] T. Sinkjrer and I. Magnussen, “Passive, intrinsic, and reflex-mediated stiffness in the ankle extensors of hemiparetic patients,” Brain, in press, 1993. [5] K. Akazawa, J. W. Aldridge, J. D. Steeves, and R. B. Stein, “Modulation of stretch reflexes during locomotion in the mesencephalic cat,” J. Physiol., vol. 329, pp. 553-567. [6] G. L. Gottlieb and G . C. Argawal, “Stretch and Hoffmann reflexes during phasic voluntary contractions of the human soleus muscle,” Electroencephal. and Clinic. Neurophysiol., vol. 44, pp. 533-561, 1978. [7] J. W. Hunter and R. E. Keamey, “Dynamics of human ankle stiffness: Variation with displacement amplitude,” J. Biomech., vol. 5, no. 10, pp. 753-756, 1982. [8] T. Sinkjaer, E. Toft, S. Andreassen, and B. C. Homemann, “Muscle stiffness in human ankle dorsiflexors: Intrinsic and reflex component,” J. Neurophysiol., vol. 60, pp. 1110-1121, 1988.
[9] L. M. Nashner, “A model describing vestibular detection of body sway motion,” Acta Otolaryng., vol. 72, pp. 429-436, 1971. [IO] J. F. Yang, R. B. Stein, and K. James, “A method to apply muscle stretch during walking in humans,” Canadian Soc. Biomech., Ottawa, Ont., Canada, vol. 5, pp. 42-43, 1988. [ l l ] R. B. Stein, J. F. Yang, M. Edamura, and C. Capaday, “Reflex modulation during normal and pathological human locomotion,” in Neurobiological Basis of Human Locomotion, 1991, pp. 335-346. [12] M. Llewellyn, A. Prochazka, and S. Vincent, “Transmissiod of human tendon jerk during stance and gait,” J. Physiol., vol. 382, p. 82P, 1987. [13] D. A. Winter, The Biomechanics and Motor Control of Human Gait. Waterloo, Canada: University of Waterloo Press, 1988. [14] V. T. b a n , H. J. Ralston, and F. Todd, Human Walking. Baltimore, MD: Williams & Wilkins, 1981. [15] M. Bendt, M. Damsgaard, W. Markus, L. Poulsen, and L. Tvermose, “Bowden cables,” Aallborg Univ., Denmark, Intem. Rep., CK-6 (3.122, 1993. [16] J. C. Houk and W. Z. Rymer, “Neural control of muscle length and tension,” in Handbook of Physiology, pp. 257-323, 1981. [I71 C. Frigo and P. Crenna, “Neural control of locomotion: Some recent advancements in methodological approach,” in Proc. Topic. Workshop: Restoration of Walking Aided by Functional Elec. Stimulation, Ja. A. Van Alster, Ed. Milan, Italy, Edizione Profucentuk, 1987. 1181 C. Capaday and R. B. Stein, “Amplitude modulation of the soleus HReflex in the human during walking and standing, The J. Neurosci., vol. 6, no. 5, pp. 1308-1313, 1986. I191 R. B. Stein and C. Capaday, “The modulation of human reflexes during functional motor tasks,” TINS, vol. 11, no. 7, pp. 328-332, 1988.
Jacob Buus Andersen was bom in h h u s , Denmark, in 1965. He received the M.Sc. degree in electronic engineering with a biomedxal engmeermg opbon from Aalborg University in 1991. Since 1992, he has been a Ph.D. student at the Center for Sensory-Motor Interactions, Department of Memcal Informatics and Image Analysis at Aalborg Umversity. His education includes a European course on biomedical engineering and medical physics in 1990 at the University of Patras, Greece, and research in cerebellar control of motor behavior in 1991 at the Department of Physiology, Northwestem University, Chicago, IL. His reseakh interests include electrophysiology, biomechanics, and human motor control.
Thomas Sinkjaer ”84) received the M.Sc.E.E. from Aalborg University, Denmark, in 1983, with specialization in biomedical engineering, and the Ph.D. degree in 1988. His dissertation was a jointventure project between the University of Calgary, Canada, and Aalborg University. From 1984 to 1986 he studied at the Department of Clinical Neurosciences, University of Calgary, and during 1989-1990 at the Department of Physiology, Northwestem University, Chicago, E. Since 1986. he has been with the Der”ent of Medical Informatics and Image Analysis as an Associate Professor h 1992 he was made Head of the Department and, most recently, Research Council Professor and Head of the Center for Sensory-Motor InteractionsDTeural Prostheses, whch was established in 1993 under the Department of Medical Informatics and Image Analysis. His research interests include electrophysiology, biomechanics, motor control (human sensory-motor interaction), and neural prostheses. Dr. S w a r holds a wide number of offices, among them an Expert for the EC @G XIU) Program Biomed I and a referee for various intemational journals.
