You are on page 1of 7

1100

Feedback of Ankle Joint Angle and Soleus Electromyography in


the Rehabilitation of Hemiplegic Gait
G. Robert Colborne, PhD, Sandra J. Olney, PhD, Malcolm P. Grifin, PhD

ABSTRACT. Colborne GR. Olney SJ, Griffin MP. Feedback of ankle joint angle and soleus electromyography in the
rehabilitation of hemiplegic gait. Arch Phys Med Rehabil 1993;74:1100-6.
l A computer-assisted feedback system was developed to present to walking subjects instantaneous feedback of their
muscle activity or joint angular excursions during gait. Targets for muscle activity or joint motion were displayed on the
feedback screen along with timing cues that prompted muscle activity or joint flesionlextension at specific times during
the gait cycle. The purpose was to compare the effectiveness of joint angle and eiectromyographic (EMG) feedback to a
focused program of physical therapy for gait. Eight hemiplegic stroke patients were treated with ankle joint angle
feedback, EMG biofeedback from the soleus muscle, and conventional physical therapy for gait in a three-period
crossover design. PT was given either first or last in the sequence of treatments. Gait analysis prior to and following each
type of treatment revealed that the feedback treatments resulted in significant increases in stride length and walking
velocity and in positive changes in push-off impulse, gait symmetry, and standing weight-bearing symmetry, as evaluated
in a general linear model and paired t-tests. Overall, physical therapy produced no significant changes. However, when
physical therapy was the first treatment of the sequence, significant increases in stride length and velocity were observed.
When physical therapy was last, there were significant negative changes in gait symmetry and standing weight-bearing
symmetry, and negative trends in stride length, walking velocity, and push-off impulse. It is concluded that computer-as-
sisted feedback is an effective tool for retraining gait in stroke patients.
16 1993 bss the American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and
Rehabilitation

The rehabilitation of gait following stroke accounts for a The feedback of muscle electrical activity (EMG) has
large proportion of the time spent in active treatment. The been the most widely investigated feedback modality in re-
importance of walking in activities of daily living makes habilitation. Using EMG biofeedback. patients have been
developing gait competence an important part of physical trained to relax hypertonic muscles7*’ or to activate flaccid
therapy. Motor reeducation can be considered from the muscles in a proximal-to-distal progression.“-” Many early
standpoint of information processing, and indeed. feedback studies were inadequately controlled. or involved anecdo-
of performance is an essential element in all contemporary tal. single-case designs, and although they showed biofeed-
models of motor learning.’ Motor program theory suggests back to be an effective form oftreatment, they did not gener-
that a program center generates a pattern of neural outflow ally prove its superiority over conventlonal physical
that causes muscles to act in a predetermined way. When therapy. More importantly, muscles have been trained out-
the program is imperfectly established. feedback of perfor- side the context of any functional activity, and although
mance is necessary, so closed-loop mechanisms dominate: isolated movements have been facilitated, deeming the
eventually, through repeated practice, the skill is learned treatment successful. little transfer has been shown to oc-
and operates with some degree of automaticity, under cur. Mulder and colleagues’.” stress the importance of
open-loop control without the use of feedback.’ The chal- learning a component movement within the framework of
lenge to therapists is to provide the patient with accurate a goal-directed action.
and immediate feedback of performance during the learn- Genu recurvatum was a focus of early applications oi
ing stages. The need to enhance this process has led to the joint angle feedback devices. which provided warning tones
development of microcomputer-based systems to monitor if a certain degree of knee extension was exceeded.13,“’
various aspects of movement, and to provide patients with Though these devices could be used during gait training,
instantaneous information about the attempted movement timing of the feedback signal to specific times in the stride
during its performance.3” cycle was not controlled. Knee flexion and ankle dorsiflex-
ion occur twice in one stride. so it is desirable to be able to
From the Dcpartmcnt ofAnatom!, (Dr. C‘olhorne). School ~I’Rchabilit~tion Ther- deliver feedback during either one. or both of these time
apy (Dr. Olnc~). Departmrnt of Mathematics and St&tics (Dr. Grdlin). Queen‘s periods.
I IniverGtv. Kingston. Ontari& Canada.
Suhmitwd for puhlicatiun October 6. IYY?. -\cccpted in ruixd tbrm Janunr~ 25. The importance of the work done by the ankle plantar-
IYYi. flexors in walking is well established. Winter” stated that
This rescnrch supported h) the National Heallh Resmrch and Dc\clopment PI-O- the plantarflexors are responsible for approximately 8097 of
gram. Health and Welfxe (‘anadn. grant 6606-2969-5 I.
No commcrclnl party having il direct or indirect intereSt in the suhjject matter ofthis the positive work done in gait. Olney and associates” rc-
XII& has confcrrcd or will confer 3 hcnrtit upon the authors or upon zmy orgnnira- ported that the profile ofwork was largely. intact in hemiple-
tlon uith \chlch the nuthon xc associated.
Rcpnnt quests IO G. Robert Colhornc. PhD. Dcpanment of Anxun~. C‘ollegc ut
gics, but that the amplitude of the positive and negative
Mcd~cine. Llmvcrsit~ of Saskatchrwan. Ssshatoon. Saskatchewan. S7N OWtI. Can- bursts were decreased. Positive and negative work at the
ads. ankle. as well as the maximum ankle plantarflexor power
L I YYi h) the 4mcr1cnn C‘ongrcss of Rehabilitation Medunc and the American
Qc;ldemy of PhYw31 Medicine and Rehahilitutlon during push-off were significantly correlated with walking
1~003-9993/“i/7311,-(1009$i.o0/0 velocity in three groups of hemiplegic stroke patients. The

Arch Phys Med Rehabil Vol 74, October 1993


BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne

ankle plantartlexors limit dorsiflexion in the early stance


phase. by contracting eccentrically, or doing negative work.
A concentric contraction follows for push-off. Generally,
when training ankle control in gait, one ofthe goals ofphysi-
cal therapy is to decrease the amount of negative work in
early stance, thus freeing the ankle to dorsiflex into a posi-
tion where the plantarflexors can contribute to forward im-
pulse through an active push-off. Despite knowing the mo-
tor objective. physical therapists have few tools to help
them teach discrete motor control, and patients receive
only manual and sub.jective. verbal feedback from the thera- Experimental design. This is a three-period crossover design, with
pist to guide them. ELGON, EMG, and PT as the three treatments. Note that PT, in
The purpose of this study was to explore the use of a its role as the control treatment, was the first treatment for four
computer-assisted feedback (CAF) system for the rehabilita- subjects, and the last treatment for the other four subjects.
tion of gait in stroke patients, and to determine whether this
method is as effective as conventional physical therapy in
developing ankle control in this group. Previous reports”.” persistence of any training effect. Each subject was ran-
have outlined the development of the CAF system. and its domly assigned to a treatment order at intake.
preliminary testing.
Gait Analysis
!CIATERIALS AND METHODS A standard procedure was followed for each gait analysis.
Subjects wore comfortable shorts and flat walking shoes.
Subjects Lower limb segments were defined by taping small reflec-
Eight volunteer hemiplegic stroke patients were selected tive markers to the centers of rotation of the hip. knee, and
to participate. All were referred from the Stroke Rehabilita- ankle joints, and to the lateral side of the foot. The land-
tion Unit at St. Mary’s of the Lake Hospital, Kingston, marks upon which these markers were placed were the
Ontario by the senior physical therapist assigned to this case greater trochanter of the femur at the hip, the lateral epi-
load. All were at least 7 months post-stroke (mean, 17 condyle of the femur at the knee. the lateral malleolus at the
months). Criteria for selection included independence in ankle. and the lateral aspects ofthe heel and fifth metatarsal
walking with or without assistive devices. ability to follow head. This allowed the lower limb to be modelled as three
simple instructions, and ability to interpret and respond to segments, consisting of a thigh. a shank. and a foot. The
the feedback signal. In addition, a common pattern of acti- head, arms, and trunk were considered one segment, and a
vation of the affected-side soleus muscle during gait was marker on the acromion at the shoulder connected this seg-
required, because this was to form the basis of treatment. In ment to the hip.
all selected subjects, the soleus activation profile was bipha- Each subject was filmed while walking along the gait
sic. The muscle became active very soon after foot contact, walkway, which extended 9m along the base of a back-
then built quickly to a high level. restraining passive dorsi- ground wall. Reference distance markers 25cm apart were
tlexion during early and mid-stance. A second peak of activ- hxed to this wall. Horizontal distance in the plane of pro-
ity was present in late stance for the push-off. Each subject gression was measured from these reference markers, and
signed an informed consent form prior to the first data col- corrected for parallax error by reference to a meter stick
lection session. held in the center of the walkway at the beginning of each
filming session. No constraints were imposed, no walking
Esperimentai Design aids were used, and each subject was instructed to walk at
A three-period crossover design was followed (fig). The his or her own comfortable speed. Prior to the collection of
experiment consisted of three types of treatment, each data in each gait analysis session, several practice walks
given twice per week for 4 weeks (eight treatments per were taken to ensure that the subjects were walking at their
block). Treatment types were ( I) EMG biofeedback from natural speed.
the affected-side soleus muscle (EMG), (2) sagittal plane A 16mm LOCAM” tine camera on a manually guided
joint angle feedback from the affected side ankle joint (EL- tracking cart recorded each walking trial from a distance of
GON), and (3) physical therapy (PT). Each of the eight 480cm lateral to the subject. Film speed was 50frames/sec-
subjects received all three treatments. but in different ond. The three middle strides of a five-stride walking trial
orders. In its role as the control treatment, PT was placed were recorded, first from the right side. then from the left.
either first (PTF) or last (PTL) in the sequence of treat- giving three continuous strides from each side. In three sepa-
ments. yielding four protocols across two sets of four sub- rate trials on each side. ground reaction force data from a
jects. triaxial force plate mounted flush in the gait walkway were
4 baseline gait analysis (GA 1) was performed prior to recorded and stored on magnetic disk.
the first treatment block, then again during the week follow- Following the walking trials. standing weight-bearing
ing each block of treatments (GA 2, 3.4). A fifth gait analy- symmetry wasassessed by having the sub,ject stand with one
sis (GA 5) was done 1 month after the fourth. with no inter- foot on the force plate, and the other offthe plate, shoulder-
vening treatment. as a follow-up test to determine the width apart. Vertical force data from each side were col-

Arch Phys Med Rehabil Vol74, October 1993


1102 BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne

lected separately for three trials of three seconds each, and tube (CRT) screen as a horizontal line that moved upwards
averaged to yield weight-bearing ratio as the percentage of and downwards with changes in signal voltage. Two ampli-
body weight borne by each foot when standing (affected tude targets could be set and displayed on the screen as
side/unaffected side). broken horizontal lines: one target could represent an up-
per threshold, used to increase EMG activity or joint flex-
Data Analysis ion, while the other could represent a lower threshold, to
Raw film data were reduced manually using a Vanguard encourage decreased EMG activity, or joint extension. The
projector and a digitizing tablet with custom software that targets could be set and used individually, or together. Each
calculated the instantaneous knee and ankle angles from target had, overlaid on itself on the right side of the screen, a
the X-Y positions of the reflective joint markers. Knee an- timing cue that would flash on the screen at a specified time
gle at ipsilateral foot contact, and maximum knee flexion after foot contact, to indicate the time in the gait cycle when
during the swing phase were recorded for both sides. Stance that target should be achieved. This time period was speci-
phase range of motion of each ankle, calculated by adding fied by on and off values signalling the beginning and end of
the maximum dorsiflexion in mid-stance and plantarflex- a “timing window” within the stride cycle. The average
ion at toe-off was likewise recorded. Stride length was deter- time of the stride cycle was determined during several base-
mined from the linear horizontal displacement of the hip line walks at the beginning of each session, using a foot-
marker in the plane of progression between ipsilateral foot switch worn under the shoe. Subjects were given verbal
contacts, referenced to the background wall markers. Stride feedback as necessary between the feedback trials to main-
time was indicated by a mechanical film frame counter, tain this cadence to ensure that the timing cues were given
and was recorded from foot contact to ipsilateral foot con- at the appropriate times in the stride cycle.
tact. Velocity was then computed as stride length divided As the subjects walked, they attended to the visual feed-
by stride time. Temporal gait symmetry was calculated as back on the CRT screen. An auditory “beep” was given if
the swing time of the affected limb divided by that of the they exceeded a target during the appropriate timing win-
unaffected limb (swing ratio). In addition to the data re- dow. and at the end of each five- to six-stride trial. a score-
duced from the film recordings, push-off impulse was com- card appeared on the computer screen showing the number
puted from the force plate recordings as the area under the of successful attempts at each target, the average peak signal
anterior-posterior force curve (force x time) during the amplitude during the timing period, and the average time
push-off phase. in the gait cycle this peak occurred. This “knowledge of
Each gait variable was analyzed separately in a general results” reinforced the visual and auditory feedback, and
linear model. The average of three repeated measures per provided the subject with a measure of performance during
subject was computed for each variable at each stage of that walking trial. These values were discussed with the sub-
assessment (GA 1 to 5). Two models were used. Included as ject at the completion of each trial, to enable the formula-
independent variables in the first model were the three tion of a strategy to meet the amplitude and/or timing tar-
treatment types (PT. ELGON, EMG) and the subject code gets during the next trial.
to account for intersubject variability. Gait analysis num-
ber was included as a covariant in preliminary analysis to EMG Biofeedback Treatments
assess the effect of time in the treatment sequence on the
Half-hour treatments were given twice weekly for 4
outcome of each dependent gait variable. If the effect of
weeks. The subjects wore their own comfortable shorts and
time was significant (p < 0. I), then the output ofthat model
walking shoes, and Beckmand EMG surface electrodes were
was reported. Otherwise, the model was run again without
applied to the skin over the affected-side soleus muscle,
the covariant. Paired t-tests were run to determine pretreat-
below the lateral gastrocnemius. The collected signal was
ment to posttreatment changes for each treatment.
bandpass filtered between frequencies of 30 and 1,OOOHz,
The second model included as independent variables the
then amplified by a gain of 1,000. The resulting signal was
three treatment types. but with PT broken into its two com-
rectified, subjected to a second-order Butterworth low-pass
ponents, PTF and PTL. The effect of time was assessed as
filter at a cut-off frequency of 3Hz, then converted to digital
outlined above, and paired t-tests likewise tested for change.
form and displayed on the CRT screen. A footswitch was
A significance level of 0.1 was chosen on the basis of the
placed under the shoe in the area of first contact with the
small sample size, which was further reduced by splitting
floor. Average stride time was computed from three base-
the PT treatment into two smaller subgroups.
line walks of five to six strides, to determine the real time
latencies for each timing window following foot contact.
Biofeedback System Three further walks were performed to collect EMG data
The biofeedback equipment comprised an IBM-compati- during the two timing windows. The first was the period
ble microcomputerb with 5 12Mb of memory and a double between 10% and 30% of the stride cycle, whereas the sec-
floppy disk drive. An analog-to-digital conversion system ond represented the push-off period, between 35% and 55%
used a Labmastef card, which accepted voltages between of the cycle. The upper target was set to a value 10% above
+ 10 and - 1OV. and custom software permitting variable the mean peak EMG value recorded during the push-off
data collection periods and sample rates. The analog signal period, and the lower target was set to a value 10% below
from any transducer could be relayed from a belt-pack the mean low peak EMG value recorded during the earlier
worn at the subject’s waist and displayed on a cathode ray timing window. Feedback trials always began with the

Arch Phys Med Rehabil VoI74, October 1993


BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne 1103

lower target alone. to train the subjects to relax their plan- values are listed irrespective of the position of the particular
tarflexors during the early stance phase, thus freeing the treatment in the treatment stream, and as such, reflect carry-
ankle joint for passive dorsiflexion during this time. Ten over from previous treatment blocks in the sequence.
walking trials were performed with this target, then it was Walking velocity improved from 0.48m/sec at baseline to
replaced on thhescreen by the upper target. set to encourage 0.6 lm/sec following the EMG biofeedback treatment, and
increased EMG activity during the late stance phase. Ten this was maintained at follow-up for an overall increase of
further walking trials were performed with this target. After 19%. Increases in stride length and decreases in stride time
demonstrated proficiency with each target separately, both both contributed to this change. Stride length increased by
targets were activated together for a final IO trials. 14cm over the baseline measure of 66cm (2 1%) and stride
time decreased by 0.13sec from its baseline of 1.44sec (9%).
ELGON Feedback Treatments Knee and ankle joint kinematics changed as a result of all
Half-hour treatments were given on the same schedule as three treatments; both the affected and unaffected sides
above. A uniaxial electrogoniometer with a parallelogram demonstrated increases in knee flexion during swing, knee
linkage was attached to the ankle using a footplate under flexion at foot contact, and ankle range of motion. The
the heel inside the shoe, and a strap around the calf. Dorsi- ELGON treatment resulted in the greatest knee joint flex-
flexion of the ankle resulted in an increase in voltage ion and ankle joint range on the affected (treated) side:
through the potentiometer, and upward movement of the maximum knee flexion during swing increased by 8” and
feedback signal on the CRT screen. Plantarflexion pro- ankle range improved by 7”. The ELGON and EMG treat-
duced downward movement of the feedback line. Follow- ments produced the greatest increases in push-off impulse
ing the recording of stride time. timing windows were set for (30%) and these were maintained through the follow-up
two periods in the gait cycle. and baseline angular data were period. Swing ratio was positively affected by the feedback
recorded for three walking trials. The first window was set treatments, and negatively by PT; a shift toward 1.00 re-
for 35’%to 45”‘rof the stride time. representing the period of flects better gait symmetry. Standing weight bearing ratio
maximum dorsiflexion during stance, whereas the second was improved after all three treatments. with the greatest
was set for the period of maximum plantarflexion during gains following ELGON and EMG.
push-oti, between 50% and 70%‘.Targets were set to encour- Table 3 lists the statistically-adjusted least square mean
age 3” more dorsiflexion and plantarflexion at the appro- values for each variable before and after each of the three
priate times in the gait cycle, and feedback beeps were given types of treatment. In each case, the least square mean
when the targets were exceeded at the proper times. value is coded as “0,” representing the measure on that
variable immediately prior to the particular treatment, and
Physical Therapy Treatments ” 1,” which represents the measure after treatment. In this
The therapy protocol focussed on facilitation of isolated way, the discrete effect of the particular treatment is demon-
and controlled use of the ankle plantarflexors during gait, strated, and the carryover effect of previous treatment
using treatments that combined the handling principles of blocks is minimized. Only two variables were significantly
the neurodevelopmental (NDT) approach” with principles affected by time in treatment: affected-side knee angle at
of motor learning. The two physical therapists involved foot contact, and maximum knee flexion in the swing
were both NDT certified, and familiar with the concepts of phase. The ELGON and EMG feedback treatments were
the Motor Relearning Program for Stroke.“’ Emphasis was associated with further significant gains on these measures.
placed on promoting dynamic alignment in stance through Significant increases in walking velocity were observed fol-
the trunk and the affected lower extremity. This sometimes lowing the ELGON and EMG treatments. These were ef-
required manual mobilization of the ankle or other lower fected primarily through increases in stride length. although
extremity joints to improve joint range. This was combined stride time was decreased through E:MG. 4ctive range of
with training of specific components of gait as well as motion of both ankles was significantly increased following
transfer of learning to the gait pattern as a whole. The spe- ELGON. in response to the dorsiflexion and plantarflexion
cific components emphasized were the reduction of triceps targets given in treatment. Push-off impulse tended to de-
surae hyperactivity in the early stance phase, improvement crease on both sides through PT, and tended to increase in
of ankle dorsiflexion range, and the development of active ELGON and EMG, though nonsignihcantly. Both swing
ankle plantartlexion in late stance. The gait components ratio and weight-bearing ratio showed positive change after
and the pattern as a whole were practiced consistently with ELGON and EMG. whereas PT produced a change in the
verbal and manual feedback from the therapist. Half-hour opposite direction.
treatments were given twice per week for 4 weeks. The pro- Table 3 lists the output of the expanded model. where PT
gram content was as standardized as possible, given inher- was broken into its two component parts, PTF and PTL.
ent differences in functional and physical abilities between Again, affected-side knee flexion at foot contact and maxi-
subjects. Two physical therapists (Martin and Arthur) mum knee flexion in swing were the only gait variables
jointly assessed each subject. They agreed on the treatment affected by time in treatment, and the posttreatment mea-
protocol and .Arthur conducted all of the PT treatments. sures were significantly greater after ELGON and EMG.
Knee flexion during swing ofthe unaffected limb was signifi-
RESULTS cantly increased after PTF. Significant increases in walking
Mean values for each of the gait variables measured in velocity were observed after PTF. ELGON and EMG, and
the gait analyses are presented in table 1. The posttreatment these were effected through increases in stride length and

Arch Phys Med Ftehabil Vol74, October 1993


1104 BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne

Table 1: Baseline, Post-treatment and Follow-up Mean Values and Standard Deviations (brackets) for Gait Variables

Baseline Post-m- Post-ELGON Post-EMG Follow-up

Velocity (m/set) 0.48 0.56 0.59 0.6 I 0.62


(0.22) (0.21) (0.21) (0.19) (0.21)
Stride length (m) ‘0.66’ 0.73 0.77 0.78 0.80
(0.27) (0.24) (0.24) (0.23) (0.25)
Stride time (set) 1.44 1.33 1.34 1.31 1.31
(0.21) (0.14) (0.11) (0.16) (0. I I )
Affected knee angle at 10.8 13.6 16.4 14.6 13.3
foot contact (degrees) (7.4) (7.0) (8.9) (8.1) (7.0)
Unaffected knee angle at 15.3 17.3 16.4 16.0 17.7
foot contact (degrees) (11.8) (8.7) (11.2) (8.4) (8.5)
Affected knee flexion in 38.9 43.3 46.4 43.7 42.8
swing (degrees) (15.2) (13.1) (13.7) (14.5) (15.3)
Unaffected knee flexion in 58.7 64.4 64.5 65.5 65.8
swing (degrees) (9.2) (5.0) (6.0) (4.5) (4.2)
Affected ankle ROM 19.7 22.2 26.6 22.3 7’ 4
__.
(degrees) (11.7) (11.2) (I 1.0) (10.0) (9.4)
Unaffected ankle ROM 20.9 22.8 25.7 24.2 26.2
(degrees) (6.5) (3.5) (3.3) (2.9) (3.2)
Affected side push-off‘ 8.25 8.60 IO.61 10.81 10.31
impulse (N . set) (5.89) (3.87) (3.72) (5.61) (5.21)
Unaffected side push-off 15.1 I 15.37 15.78 17.42 17.23
impulse (N . set) (4.86) (6.1 I) (4.68) (5.05) (5.17)
Swine ratio (affected/ 1.33 I .39 I .24 1.26 1.26
unaffected) (0.50) (0.42) (0.32) (0.27) (0.27)
Weight ratio (affected/ 0.66 0.73 0.77 0.78 0.74
unaffected) (0.18) (0.14) (0.16) (0.14) (0.19)

decreases in stride time. ELGON improved the active feedback of joint angle or EMG can produce measurable
ranges of motion ofboth the affected and unaffected ankles. changes in gait function.
The greatest positive change in push-off impulse on the af- Only two gait variables, angle of the affected-side knee at
fected side was observed following EMG, although this was foot contact and maximum knee flexion during swing, were
not significant. Both PTF and ELGON were associated significantly affected by time in treatment. This indicates
with positive change on this variable, whereas PTL was as- that knee function developed in time as other aspects of gait
sociated with change in the opposite direction. Push-off im- function were improved. Given that knee control was not
pulse on the unaffected side was significantly improved by directly trained by either ofthe two biofeedback treatments,
EMG. Significant negative changes in swing ratio and it follows that return of knee function developed in parallel
weight-bearing ratio were observed following PTL, whereas with overall gait competency. Because knee kinematics
the other three treatments tended to improve these mea- were the only gait variables that were positively affected by
sures. PTL (table 3), the incremental increases in performance
measured for these variables across the three treatment
DISCUSSION blocks therefore implicates time as an effect, although when
The feedback treatment protocol developed for this study time in treatment was accounted for in the models, EL-
is substantially more complex than systems used in pre- GON and EMG also produced significant change. Olney
vious reported studies. Other investigators’3,‘4 isolated genu and associates’6 showed a strong positive relationship be-
recurvatum as a gait deficit that could be treated through tween walking velocity and affected-side knee flexion in
the use of a feedback tone when the knee passed into hyper- hemiplegics. This was likely a function of stride length. Al-
extension. Their equipment and methods did not incorpo- though stride length was not reported by Olney,” there was
rate the use of timing cues for the proper knee motion, but a tendency toward greater hip extension at the end of stance
this was not important, because they were only attempting and a larger burst of work by the hip flexors at the initiation
to prevent knee hyperextension at all points in the gait cy- of swing as velocity increased. The angular rotation of the
cle. The EMG feedback literature is dominated by reports thigh produced by this burst of positive work would transfer
of the use of EMG biofeedback to decrease reflex responses energy to the shank during this period, and would likely
to imposed muscle stretch and its use to produce increased influence passive knee flexion in the process.
muscle activity in isolated movements. It has been shown, Separating the effects of physical therapy into changes
in both of these cases, that transfer of training to functional, produced by PTF versus those produced by PTL serves to
goal-directed activities is poor, and that it is important to illustrate the relative efficacy ofphysical therapy in compari-
train components of a movement within the context of the son to the two feedback treatments: however, it is fair to
overall movement pattern.12 The computer-assisted feed- examine the overall effect of PT regardless of its position in
back system used here was developed in response to these the treatment stream, because ELGON and EMG were ex-
shortcomings. The encouraging results of this study indi- amined in this way. Table 2 shows that none of the changes
cate that focussed treatment of a specific gait deficit using resulting from PT were significant, although in most cases

Arch Phys Med Rehabil Vol74, October 1993


BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne 1105

Table 2: Least Square Mean Values for Pretreatment (0) The tone provided a discrete “hit” message when the targets
and Postreatment (1) Measures on Gait Variables were exceeded. and the visual display provided additional
_..- _____-
knowledge of results by informing the subject by how much
PI EI,GON EM<;
the targets were either overshot, or undershot within that
0 0.55 I 0.534 stride. At the end of each walking trial. the scorecard pro-
I 0.564 0.58 I+ vided further information concerning the average response
Strdc length (m 1 0 0.729 0.710
amplitude over that trial, and this information would im-
I 0.747 0.7hhf
Stride time (WC) 0 I.379 I .36Y pact on subsequent trials.
I I.339 I .34x The other mechanism is “feed forward.” whereby success
4tFected knee angle at foc,t 0 t I.09 10.9X or failure at reaching a target during one stride shapes the
contact (degrees)* I 13.6’) 14.70+ motor output in subsequent responses. This follows Mar-
I InaHected knee angle at foot 0 16.13 15.39 17.06
t
teniuk’s” model, in that each attempt to produce the pre-
contact (degrees) 16.72 17.45 15.7’)
4tfected knee tlexion In 0 10.67 39.3’) 39.58 scribed movement provides knowledge of results in the
swing (depree:;~* I 13.06 34.30’ 44. If? form of “hit” or “no hit.” and this information is used to
I inafected hnec Ilexlon in 0 62.34 62. I3 62.70 construct the conceptual image ofthe subsequent response.
cwing (degrees) I 64.04 64.75 64. I7 which then runs largely under open-loop control. Over re-
Altitctcd x~kle range of 0 x.64 70.87 ‘2.50
mcltlon (degree\) I 32.18 74.0 I + 37 71
_-._ _ peated attempts (assuming the capacity for change exists).
I InuH&ted ankle range ot 0 23.32 ‘3 ‘6
__.- -_.
‘? 92 the response movement converges with the conceptual
motion (degrees) I 23.76 74.82’ 24.16 image of the desired response.
Alfected side nush-all 0 IO.07 9.21 x.74 The subjects reported that the auditor! feedback tone
impulse (N. sm.) I 0.09 0.92 IO.41
0
was of more use to them than the continuous visual feed-
I’natltcted side push-oil 16.23 IS,Y3 15.10
impulse (N .scc) I 15.X1 16.12 I6.94+ back. Many of them had become reliant on watching their
Swing ratio (atii:cte& 0 I.15 I .35 I.37 feet as they walked. and chose to concentrate their attention
unaH‘ected) I 1.35 I.15 1.2X on the auditory tone and the motor output necessary to
height mtlo (;111cctc’d/ 0 0.734 0.703 0.692 achieve the target. They were coached to watch the feed-
unafi‘ectcd) t 0.7 I7 0.747 0.759
back display, however, and the concomitant adjustment in
PI F and PTI. are combined as P’I. posture likely affected their gait in a positive way.
* FHbct of time in trcatmcnt was signiticant 11 < 0.05. Finally. the persistence of training through the l-month
+p ’ 0 I, follow-up period indicates that the subjects incorporated
i 0 ’ O.fIS.

Table 3: Least Square Mean Values for Pretreatment (0)


the changes in function were in the same direction as those and Posttreatment (1) Measures on Gait Variables
produced by the other treatments. Notable exceptions were ___-
push-off impulse, swing ratio. and standing weight-bearing PfP I<:I.C;ON EiXlG Pm.
__________ ___ ____
ratio. The first two are doubtless connected. because in- 0
Velocity (m/xc) (1.52X 0.53X 0.512 0.5hY
creasing the push-off activity of the affected side should I 0.54? 0 5R3t 0.59x* 0.55I
prolong the stance phase. and propel the affected limb Stride length t m) 0 0.7 I I 0 71h 0.7 I h 0.7-IY
faster through the swing phase. However. despite an aver- I 0.775’ 0 7:lt 0.770* 0.73X
age increase in velocity. push-off impulse was decreased, Stride time (xc) (I I .-hoi; I 354 I.?XI I ..Xl
I I .IXJ~ I 33h I .:0X’ I .34Cl
suggesting that lift-off of the foot occurred through active 4ttected knee angle 31 fhot 0 12.33 I I .K’ I I.‘,3 I3.10
hip flexion. and indeed. this is a valuable mechanism of contact (degrees)* t t4.75 15.3’l’ t 5.2h’ IJ.09
positive work at the end of stance in cases where ankle plan- Unaffected knee angle at 0 17.17 IS.‘! 17.42 If,.13
tartlexion power is poor.” This is supported by the observa- foot contact (degrees) I 16.35 t 7.x; t h.-‘O 17.4’)
Afected knee tleuion in 0 40.X3 Jo.tAJ 1O.hX 42.50
tion of increased knee flexion in swing. I &LX0
swing (degrees)* JJ.YJ’ 44.--l+ 43. I-1
The mechanism of learning in this experiment deserves llnaffected knee flexion In 0 hl .X6 h2.M h2.cr.3 hI.72
some discussion. The term “biofeedback” implies a closed- swing (degrees) t hS.tiX’ h4.X’~ h4.h I f,l.(li
loop system, and as such, the continuous visual feedback RH‘ectcd anhle range 01 II 20.h7 30.-13 27. ! t 73.12
would be used during the actual production of the response. motion (degrees) t 13.29 3.3w+ ?l.Sl 20.x I
Irnaflectcd ankle range ot 0 ‘2.-1X 22:3x 23.05 77.9h
Wetzel and German” reported that 1 IOmsec was the short- motion (degrees) I 24.77 24.X” 34-TI 23.30
est latency observed for able-bodied subjects to respond Affected side push&’ 0 X.74 b..(1
7 x. ix ‘I.LJI
with EMG activity to a discriminative stimulus during slow impulse (N . see) I 93.7 II 7’1 4.70 X.Oh
walking at 0.9m/sec. In the present study. each of the tim- II naffected side push-oil 0 Ih.-lO I i\.lS I-b.‘)7 1.57
impulse (N. SK) I 15.32 Ii.% Ih.SO+ 15.x5
ing windows were active for 20% of the stride cycle, for an 0 t 47 I .a1 I ix I ./5
Swing ratio (atTected/
average time of 790msec. Whether these hemiplegic sub- unaffected I I 131 I .?’ I. :s I .-1?
jects had time to adjust their EMG activity in response to Weight ratio (Sected/ 0 0.664 O.hY(l 0.07x 0.750
the visual signal within the duration of one timing window unaffected) I 0.755 I).730 0. ?J I O.f>70’

is speculative, and so it is unknown whether the continuous


PT IS hrokcn into its two components. PI’F and Pl 1.
visual feed’back of joint angle or EMG played any role * Et%ct of time in treatment was signihcant ,t c (I. I
beyond reinforcing the auditory feedback tone that indi- ‘/I < 0.1.
cated a successful attempt to meet the amplitude targets. t ,’ i o.oi.

Arch Phys Med Rehabil VoI74, October 1993


1106 BIOFEEDBACK OF JOINT ANGLE AND EMG, Colborne

these changes into their walking behavior. It is possible that cle contracture in children with cerebral palsy. Dev Med Child Neural
the treatments decreased the energy cost ofwalking in these 1989:3 I:47 l-80.
9. Middaugh SJ. Miller MC;. Electromyographic fcedhack: effect on vol-
subjects: this would be self-reinforcing. Though it is widely untary muscle contractions in paretic subjects. .Arch Phys Med Reha-
believed that each of us chooses a gait that reduces our bil I980:6 124-9.
energy costs to their lowest levels per unit of distance IO. lnglis J. Donald MW. Monga TN. Sproule M, Young MJ. Electromyo-
walked, in the pathological case, compensations are made graphic biofeedback and physical therapy of the hemiplegic upper
limb. Arch Phys Mcd Rehabil 1984:65:755-9.
that are not necessarily energy efficient.24 The need for sta-
I I. Wolf SL. Binder-Macleod SA. Electromyographic biofeedback appli-
bility and confidence m the overall pattern may override cations to the hemiplegic patient: changes in lower extremity neuro-
efficiency, and lead to a walking pattern that provides the muscular and functional status. Physical Therapy l983;63: 1404-I 3.
patient with a safe gait, although energy consuming. The 12. Mulder T. Hulstijn W. Van der Meer J. EMG feedback and therestora-
psychological impact of improving gait beyond a long-es- tion of motor control. .4m J Phys Med 1986:65:173-8X.
13. Koheil R. Mandel A. Joint position biofeedback facilitation of physi-
tablished compensatory pattern may have sensitized the cal therapy in gait training. Am J Phys Med 1980:59:288-97.
subjects to the nature of their energy-consuming compen- 14. Hoque RE, McCandlcss S. Genu recurvatum: auditory biofeedback
sations, and motivated them to improve further. treatment for adult patients with stroke or head injuries. Arch Phys
Med Rehabil 1983:64:36X-70.
15. Winter DA. Energy generation and absorption at the ankle and knee
Acknowledgments: The authors gratefully acknowledge the collabora-
during fast. natural. and slow cadences. Clin Orthop Rel Res
tion of staff and patients at St. Mary’s of the Lake Hospital. Kingston.
1983: 175: 147-54.
Ontario. and particularly the assistance of Cally Martin. BSc(PT) in the
16. Olney SJ. Griffin MP. Monga TN. McBride ID. Work and power in
assessment and selection of subjects and the design of the physical therapy
gait of stroke patients. Arch Phys Med Rehabil 199 1:72:309- 14.
protocol. Lin Arthur. BSc(PT) is thanked for her participation in the de-
17. Colborne CR. Olney SJ. Feedback ofjoint angle and EMG in gait of
sign and delivery of the individual PT treatments. Patrick Costigan, MSc,
able-bodied subjects. Arch Phys Med Rehabil I990:7 1:47X-83.
developed the CAF software. and Ian MacBride. MSc. assisted in the gait
18. Olnev SJ. Colborne GR. Martin CS. Joint angle feedback and biome-
analyses and reduction ofdata. Queen’s University STATLAB is acknowl-
chanical gait analysis in stroke: a case report. Physical Therap)
edged for assistance in statistical analysis. 1989;69:863-70.
19. Bobath B. Adult hemiplcgia: evaluation and treatment. 2nd rd. Lon-
References don: William Heinemann Medical. 1978.
I. Mulder T. Hulstyn W. Sensory feedback therapy and theoretical 20. Carr JH. Sheppard RB. A motor relearning programme for stroke.
knowledge of motor control and learning. Am J Phys Med London: William Heinemann Medical. 1982.
I984:63:226-44. 21. Winter DA, Olnep SJ, Conrad J. Ounpuu S. Gage JR. Adaptability of
2. Keele SW. Summers JJ. The structure of motor programs. In: Stel- motor patterns in pathological gait. In: Winters JM, Woo SL. eds.
math GE. ed. Motor control: issues and trends. New York: Academic. Multiple muscle systems: hiomechanics and movement organization.
1976:109-41. New York Springer-Verlag. 1990.
3. Shumway-Cook A. Anson D. Hailer S. Postural sway biofeedback: its 22. Wetzel MC. German Lti. Learning and locomotor reaction times.
effect on reestablishing stance stability in hemiplegic patients. Arch Human Movement Science I986:5:75- 100.
Phvs Med Rehabil I988:69:395-400. 23. Marteniuk RG. Motor skill performance and learning: considerations
4. Hiiokawa S. Matsumura K. Biofeedback gait training system for tem- for rehabilitation. Physiotherapy Canada l979:3 I : 187-202.
poral and distance factors. Med Biol Eng Comput 1989:27:8-13. 14. Olney SJ. Monga TN. Costigan PA. Mechanical energy ofwalhing of
Gahery Y. Allovon E. Blanc C. .Ambulatory computer system for re- stroke patients. Arch Phys Med Rehabil 1986:67:92-8.
5.
cording and controlling locomotion. Int J Biomed Comput Suppliers
198X:23:209-20, a. LOC\M model 5 I. Redlake Corporation. I7 I I Dell Avenue. Camp-
6. Mandel AR. Nymark JR. Balmer SJ. Grinnell DM. O’Riain MD. hrll. CA 9500X.
Electromyographic versusrhythmicpositional biofeedback in comput- b. Zenith Data Systems. model 150. 1020 Islington .4venue. Toronto.
erized gait retraining with stroke patients. Arch Phys Med Rehabil Ontario. Canada, M8Z 5X5.
I990:7 11649-54. c. Labmaster DMA. Scientific Solutions. Incorporated. 6125 Cochrane
7. Neilson PD. McCaughey J. Self-regulation of spasm and spasticity in Road. Solon. OH 44139-3377.
cerebral palsy. J Neural Neurosurg Psychiat 1982;45:320-30. d. Beckman Biopotential Electrodes. Summit Technologies. 3333 Wye-
8. Nash J. Neilson PD. O’Dwyer NJ. Reducing spasticity to control mus- croft Road. Oakville. Ontario, Canada. L6L 6L4.

Arch Phys Med Rehabil Vol74, October 1993

You might also like