Joint angle feedback and biomechanical gait analysis in stroke patients: a case report.

Joint Angle Feedback and Biomechanical Gait Analysis in Stroke Patients: A Case Report

The purpose of this case report is to demonstrate the use of gait analysis and computer-assisted visual and auditory feedback (CAF) in the treatment of a patient with stroke secondary to hemiplegia. Two-dimensional kinematic and kinetic gait analyses were performed to determine the objectives of treatment and to evaluate the results. The CAF hardware and software permitted immediate visual feedback of performance relative to the desired target with auditory reinforcement if the target was reached in the desired phase of the gait cycle. The objective of the treatment was to increase knee flexion during push-off and pull-off. After four weekly treatments, the patient's gait velocity and stride length, in addition to transfers between kinetic energy and potential energy, were significantly improved. The patient demonstrated a downward trend in total mechanical energy costs and an upward trend in the amount of energy conservation, as well as increased knee flexion. These promising results indicate a need for further research into the use of CAF as a tool in motor learning for rehabilitation.

Key Words: Biomechanics; Feedback; Gait training; Hemiplegia, gait.

For several decades it has been recognized that instrumented gait analysis can assist in the understanding and treatment of gait disorders. Devices to measure temporal characteristics such as velocity, cadence, and stride characteristics have been widely used,[1] and electromyography, frequently combined with still-frame photography[2] or electrogoniometry,[3] has been implemented to a lesser extent. It has recently become possible to use several more powerful techniques in the analyses of gait,[4-7] which proponents claim provide additional information related to the causes of the defects. Mechanical energy and power analyses are in this category. Although these methods have shown promising results in the analyses of pathological gait,[5,6,8] the use of these techniques to prescribe and evaluate treatment is still in its infancy.

A major focus in the rehabilitation of stroke patients is the improvement of their gait. Although ambulatory function is usually recovered to some degree, residual gait defects cause inefficient movement patterns and result in high energy costs of walking.[9] Learning to optimize the movements required for an efficient gait can be hindered by sensory impairment. Sensory feedback, central to any model of learning, is a necessary element in neuromuscular reeducation, and various techniques using artificial feedback have shown promise for teaching control of movement.[10,11] Patients are often trained to perform isolated movements outside of any functional context, however,[12] and such movements thus may not be incorporated into goal-directed action.

Electromyographic biofeedback has been the most popular form of experimental feedback treatment, although limitations have been identified.[13] Typically, the amplified EMG signal from a weak or paretic muscle is presented in auditory or visual form to the patient, who uses the feedback signal to learn discriminative control of the muscle. This technique can be used to facilitate relaxation of hypertonic muscle groups and activation of muscles previously assumed to be paralyzed. Surprisingly, input variables other than EMG activity have not been exploited to the same extent. For example, relatively few studies have reported on the use of kinematic feedback. A few researchers have used joint angle feedback to correct knee hyperextension with favorable results,[14,16] and feedback of sway characteristics has recently been used to reduce sway displacement and to improve loading of the affected leg.[17] It is equally surprising that the versatility, consistency, and low cost of computer systems has rarely been exploited. In an attempt to address some of these shortcomings, a computerized feedback system was developed to provide immediate, objective information and to assess progress during treatment. The purpose of this case report is to demonstrate the use of gait analysis and computer-assisted visual and auditory feedback (CAF) of specific gait variables in the improvement of gait in a patient with stroke secondary to hemiplegia.

Methods and Procedures

Patient

"Mrs P" was a 53-year-old woman who had suffered a right cerebrovascular accident, resulting in left hemiparesis, about one year prior to the study. She was hospitalized in an acute-care setting for 3 weeks and in a stroke rehabilitation unit for an additional 10 weeks, and she returned to the clinic for weekly outpatient physical therapy for 4 weeks following discharge. At the time of discharge from outpatient rehabilitation, Mrs P was walking independently without an ambulatory aid in the home, although she used a straight cane outdoors. She tended to walk with a slow and hesitant gait.

Nine months postdischarge, Mrs P was selected for feedback treatment for numerous reasons. She had observed that her gait had deteriorated, and timed tests supported this observation. She was an intelligent woman with good attention and concentration abilities. Highly motivated and cognitively unimpaired by the stroke, Mrs P was able to compensate for her residual perceptual and sensory impairment. She was well enough to tolerate attending the Motion Laboratory at Queen's University.

A pretreatment gait analysis was performed and was followed by CAF treatments once a week for a total of four weeks. A posttreatment gait assessment was performed during the week following the last treatment.

Methods of Gait Analysis

Gait assessment consisted of a two-dimensional, sagittal-plane cinematographic, kinematic, and kinetic analyses of the patient's usual gait pattern, using a LoCam camera(*) and methods described elsewhere.[6,9] The body was modeled as seven segments for kinematic and mechanical energy analyses: two feet, two lower legs, two thighs, and a single head-arms-trunk segment. Kinematic and mechanical energy analyses were performed using a Hewlett-Packard desktop computer([single dagger]) and computer programs described by Winter,[18] to obtain kinematic, mechanical energy, and joint power values for each frame of the film. The methods of obtaining kinematic, mechanical energy, and power information and descriptions of normal profiles have been reported,[18] and the application of mechanical energy and power analyses to pathological gait have appeared in recent publications.[6,8,9]

In brief, the energy of one body part at any instant in time is the sum of its potential energy, which is based on its mass and height above the ground and its kinetic energy, which comprises both translational and rotational parts. If there is an increase in either kinetic or potential energy of one segment of the body and a simultaneous decrease in the other form of energy over a short period of time, energy savings result, which are referred to as within-segment exchanges. The proportion of savings is called within-segment conservation. Transfers of energy also occur between segments, which result in between-segment conservation. The total body energy at any instant in time is the sum of the energy of each of the body segments. The total energy cost over the period of a gait cycle is the sum of all the rises and falls that occur during the gait cycle.

The net work accomplished by the muscles across the major joints can also be determined. First, the power at each instant in the gait cycle is determined as the product of the net moment and the net angular velocity of the adjacent segments.[19] Power is generated when a concentric contraction is being performed and increases with the speed of the motion and the size of the moment. Similarly, absorption occurs when an eccentric contraction is performed. The total work performed over that time period is the product of the power and the time taken to complete the gait cycle. Analyses of normal walking (Fig. 1) have shown that the major producers of positive work are the ankle plantar-flexor muscles during push-off[19] and that lesser amounts are supplied by the hip flexor muscles at pull-off (pre-swing and initial swing) and by the hip extensor muscles at the beginning of stance.[7] These bursts are shown in Figure 1 as A2, H3, and H4, respectively.

The results of three strides from each side of the patient's body were processed and plotted. The total body energy and the savings accrued as a result of between-segment and within-segment exchanges were calculated, and the savings were expressed as conservation. The total positive work performed by each of the joints was calculated from the power data.

Biofeedback Methods

The method of selecting the treatment is described in the Results section. A physical therapist (CSM) and an anatomist (GRC) specializing in feedback therapy (the feedback coach) conducted the sessions. The aim of treatment was to increase the knee's range of motion as it moved from late stance through the swing phase of gait. The signal from a uniaxial electrogoniometer([double dagger]) affixed to the patient's knee was converted to digital form with a LabMaster digital converter([section]) and calibrated in degrees, and baseline walking values were established (Fig. 2). The signal was displayed on the monitor of a desktop computer([~~]) as a horizontal white line that moved upward during flexion of the knee and downward during extension. A target for knee flexion was also displayed on the screen as a stationary horizontal line. Once determined, the patient's stride cadence was indicated by a white square in the upper left corner of the screen, which flashed on once per stride, indicating when left heel contact should occur. A red light display mounted on top of the monitor indicated a time range when maximum knee flexion should occur within each stride. If the patient reached the target within the time indicated by the red light display during training trials, she received auditory feedback. At the end of each six-stride trial, a scorecard display appeared on the screen showing the average maximum knee flexion during the prescribed time period, the standard deviation of that measurement, and the number of times the subject was successful in achieving the target value. A target value of 5 degrees over the average maximum degrees of flexion obtained during a baseline walk was selected as a training angle. We instructed the patient that a sound would be heard each time she reached the target value within the prescribed phase of the gait cycle and that the feedback coach would give her a score after each six-stride walk. After a few trials, Mrs P was able to obtain the beep with each stride, but she accomplished it by using an exaggerated flexion synergy, elevating her pelvis, and flexing her hip to accomplish the knee flexion. The physical therapist instructed her to allow her pelvis to drop rather than to elevate it and to lift her heel and allow her knee to bend before her forefoot left the ground. She was told that the beep should occur soon after this movement. The target value was decreased from 5 degrees to 3 degrees over the average maximum degrees of flexion to make the task easier. She did not use the abnormal pattern again after this instruction. Mrs P returned for a one-hour session a week later, at which time the target value was again set at 3 degrees above her baseline value for that day. At the third and fourth sessions, the target value was increased to 5 degrees over the starting value for that day. Mrs P was given no feedback at home, but she was instructed to try to repeat what she had learned in the training sessions. The patient's degree of success in reaching the target value during each trial and each session were plotted and analyzed.

Data Analysis

To test for differences between gait analysis trials, the six variables arising from temporal and mechanical energy analyses (Tab. 1) were subjected to Student's t tests. Differences with less than a 5% probability of occurring by chance alone were judged to be significant. The Type I error rate was adjusted for multiple comparisons using the relationship[20]

[a.sub.c] = [(1 - a).sup.N] when a is the probability of making a Type I error if one comparison is made and [a.sub.c] is the adjusted probability when N comparisons are performed. The remaining six variables derived from the mechanical power analyses (Tab. 2) were not subjected to statistical tests because, having been derived from only one side of the body, only three values were present in each group.

Results

Pretreatment Gait Analysis and Selection of Treatment

Table 1 includes pretreatment temporal and mechanical energy variables from six strides of the affected and unaffected sides of the body for the three gait trials. Table 2 contains descriptive statistics for kinematic variables of the knee and positive work of the ankle of the affected limb for the three trials. Representative joint angle excursions for the pretreatment measurements are shown in Figure 3. Representative joint power curves are presented in Figure 4.

Table 1 shows that the stride length was short and that the velocity was low. The mean mechanical energy costs per unit of distance walked were within the range reported for stroke patients[9] but exceeded the costs predicted for normal slow walking[21] by about 80%. Figure 3 shows that initial contact was made with the affected knee in 30 degrees or more of flexion and that the total flexion ROM was only about 12 degrees. Figure 4, when compared with the normal values of Figure 1, shows extremely low levels of power generation from both sides, with small amounts of positive work being produced by the hip flexor muscles of both sides, the knee extensor muscles of the unaffected side, and the ankles of both sides.

These results are consistent with clinical findings. Pelvic retraction of the affected side was present throughout the gait cycle. Mrs P was unable to shift her pelvis forward over her foot; she did not fully extend her hip during stance and spent a minimal amount of time in single-limb support. During the swing phase, she elevated her pelvis and swung her lower limb through with mild circumduction at the hip and an extended knee.

Although this patient clearly suffered from many gait abnormalities, her major biomechanical problem was inadequate power generation from all sources. As indicated previously, the most important sources of power generation are the ankle plantar-flexor muscles at push-off and the hip flexor muscles at pull-off. Linked to the patient's low power generation were her very limited angular changes at the hip, knee, and ankle. The limited angular changes at these joints not only reduced the level of power generation but also resulted in excessive lifting of the body, which was responsible in part for the higher energy costs. Because knee flexion must occur simultaneously with both ankle plantar flexion and hip flexion, increasing its ROM during late stance through the swing phase was seen as a key element in achieving increased power. Accordingly, the treatment plan described above was selected.

Results During Treatment

Figure 5 shows a substantial increase in the patient's success rate in reaching the target knee flexion angle for treatment sessions 3 and 4 and a decrease in the average overshoot or undershoot of the target value. In trials 1 and 2, the target value was 3 degrees over the average knee flexion angle of the baseline trials. The first treatment period was not as long as the later treatment periods, but Mrs P was quite successful in achieving the target value and in increasing the number of times she reached it. She reached the target value 55% of the time, with an average of just over 4 degrees in excess of the target angle. The second treatment period was variable. The target was reached 51% of the time, with an average overshoot of 3 degrees. In the third treatment period, although the target value was 5 degrees over the baseline value, the patient quickly learned to perform the task. She reached the target 68% of the time, exceeding it by an average of 2 degrees. During the fourth treatment, the target was reached 67% of the time, with an average overshoot of 2 degrees.

Comparison of Pretreatment and Posttreatment Gait Analyses

Table 1 contains the statistics for velocity, stride length, and energy variables from the gait analyses. After treatment, both the velocity and the stride length had increased significantly (p [is less than] .05). Average energy cost decreased from 1.58 J/kg.m before treatment and to 1.33 J/kg.m after treatment, although this was not a statistically significant change. The percentage of within-segment conservation was significantly improved (p [is less than] .05) after treatment.

Table 2 contains the knee angle and ankle work values before and after treatment. Statistical tests were not performed for the reason indicated previously. There was an increase of about 9 degrees in knee flexion and a gain in overall knee range. The total amount of work performed by the affected ankle was about twice as much after treatment than before treatment, and it accounted for a somewhat larger proportion of the total amount of work performed by the affected limb.

Discussion and Conclusions

The results of the use of gait analysis and CAF treatments were positive. Mrs P reported that she felt better walking "the new way" and that she was aware that it was both easier and faster. Her speed of walking, however, was still very low in functional terms. Mrs P's perception of feeling better may have been linked to the lower mechanical energy costs that were recorded. If this change were reflected in metabolic costs, it would give powerful reinforcement to learning the new motor pattern.

The increase in within-segment conservation that occurred as a result of treatment was a positive finding. This conservation is a result of trade-offs between height and velocity, mostly of the trunk, and has been cited as having potential importance in training.[6] Further exploration of means of exploiting this important physical phenomenon is warranted. We were also pleased with the speed with which Mrs P was able to learn the new gait pattern and the probable role that the CAF treatments played. Several factors were helpful: Mrs P was an intelligent and highly motivated patient who was able to achieve the desired movement in isolation, although not in gait. Despite variations in modern theories of motor control and motor learning,[12] information about performance is essential for the learning of a motor response.[22-24] Using the information-processing paradigm, current knowledge indicates that feedback is essential for ordering "program" changes, for monitoring errors, and for enhancing corrections. In treatment, there is an attempt to make program changes to obtain what is perceived by both the patient and the therapist to be a more normal gait. Feedback is thus of paramount importance in making these changes, and the addition of artificial forms of feedback to deficient forms should be expected to assist the speed of learning.

Several kinds of feedback were discussed in this case report, and it was clear to the feedback therapist during treatment sessions that the patient selected from among them. Immediate visual feedback moved continuously up and down the screen. Immediate discrete auditory feedback was provided by the sound indicating that the target value had been reached. Delayed visual feedback was provided by the scorecard, and delayed auditory feedback was given when the feedback coach verbally reported the results to the patient. Mrs P seemed to make the most consistent use of the immediate discrete auditory feedback, possibly because she did not feel secure while watching the screen. She reported that the consistency and immediacy of the feedback were very helpful to her and also gave her a sense of control over her own learning.

Despite the probable value of the CAF treatments, it was not possible initially to obtain the desired movement without the additional input by the therapist, probably because it was possible to achieve the movement in several different ways. Other users of feedback have also used it only in conjunction with other forms of treatment.[25] Although it is possible to increase the domain of motion in the feedback by combining two or more signals in a single display, we expect that the kind of verbal instruction provided by the therapist in this case study will usually be required in conjunction with feedback.

This case report did not address the important issue of persistence of learning. Mrs P was encouraged to try to practice at home what she had "learned" in the laboratory, without knowing whether it would be possible. Mrs P reported that she rehearsed the training sessions as she walked at home and that she "heard the beeper in her head." That is, during the training sessions she probably formed associations with other sensory inputs--likely proprioceptive and visual--that allowed her to walk without feedback as she had with feedback during the final gait analysis. A great deal more research on the retention of feedback-learned behavior is needed.

A final comment concerns the value of these complex and expensive gait analyses and the practicality of their routine use. In short, their value depends on the proposed use of the information. In determining treatment, our limited experience with mechanical energy and power analyses[6] suggests that the treatment emphasis is influenced a great deal by information from the biomechanical analyses. Similarly, if it is important to know the precise nature of changes that occur as a result of treatment, a full analysis is required. If the aim is simply to monitor the patient's progress, however, it could be sufficient to measure a few important variables such as velocity and knee angle. Currently, the analysis system is probably not practical for widespread routine use, but technological development may make such systems much less labor intensive and therefore much more feasible for clinical use during the next decade. [Figure 1 to 5 Omitted] [Tabular Data 1 to 2 Omitted]

(*)Model 51, Redlake Corp, 1711 Dell Ave, Campbell, CA 95008. ([single dagger])Model 9845, Hewlett-Packard (Canada) Ltd, 2670 Queensview Dr, Ottawa, Ontario, Canada K2B 8K1. ([double dagger])Physio-Goniometer, Royal Regional Rehabilitation Centre, 505Smyth Rd, Ottawa, Ontario, Canada K1H 8M2. ([section])Scientific Solutions, Inc, 6225 Cochran Rd, Solon, OH 44139. ([~~])Model 150, Zenith Data Systems, 1020 Islington Ave, Toronto, Ontario, Canada M8Z 5X5.

References

[1]Perry J, Bontrager E: Development of a gait analyzer for clinical use. Transactions of the Orthopaedic Research Society 2:48, 1977. [2]Knutsson E, Richards C: Different types of disturbed motor control in gait of hemiparetic patients. Brain 102:405-430, 1979. [3]Peat M, Dubo HI, Winter DA, et al: Electromyographic temporal analysis of gait: Hemiplegic locomotion. Arch Phys Med Rehabil 57:421-425, 1976. [4]Winter DA: Biomechanical motor patterns in normal walking. Journal of Motor Behavior 15:302-330, 1983. [5]Olney SJ, Costigan PA, Hedden DM: Mechanical energy patterns in gait of cerebral palsied children with hemiplegia. Phys Ther 67:1348-1354, 1987. [6]Olney SJ, Jackson VG, George SR: Gait reeducation guidelines for stroke patients with hemiplegia using mechanical energy and power analyses. Physiotherapy Canada 40:242-248, 1988. [7]Winter DA: The Biomechanics and Motor Control of Human Gait. Waterloo, Ontario, Canada, University of Waterloo Press, 1987. [8]Winter DA: Concerning the scientific basis for the diagnosis of pathological gait and for rehabilitation protocols. Physiotherapy Canada 37:245-252, 1985. [9]Olney SJ, Monga TN, Costigan PA: Mechanical energy of walking of stroke patients. Arch Phys Med Rehabil 67:92-98, 1986. [10]Basmajian JV: Biofeedback in rehabilitation: A review of principles and practices. Arch Phys Med Rehabil 62:469-475, 1981. [11]Marzuk PM: Biofeedback for neuromuscular disorders. Ann Intern Med 102:854-858, 1980. [12]Mulder T, Hulstyn W: Sensory feedback and theoretical knowledge of motor control and learning. Am J Phys Med 63:226-244, 1984. [13]Wolf SL: Electromyographic biofeedback applications to stroke patients: A critical review. Phys Ther 63:1448-1459, 1983. [14]Koheil R, Mandel AR: Joint position biofeedback facilitation of physical therapy in gait training. Am J Phys Med 59:288-297, 1980. [15]Hogue RE, McCandless S: Genu recurvatum: Auditory biofeedback treatment for adult patients with stroke or head injuries. Arch Phys Med Rehabil 64:368-370, 1983. [16]Fenton RM, Flowers WC: Use of a joint angle biofeedback system in gait training of stroke patients. In Stein JL (ed): Biomechanics of Normal and Prosthetic Gait. New York, NY, American Society of Mechanical Engineers, 1987, pp 83-86. [17]Shumway-Cook A, Anson D, Haller S: Postural biofeedback: Its effect on reestablishing stance stability in hemiplegic patients. Arch Phys Med Rehabil 69:395-400, 1988. [18]Winter DA: Biomechanics of Human Movement. New York, NY, John Wiley & Sons Inc, 1979. [19]Winter DA: Energy generation and absorption at the ankle and knee during fast, natural and slow cadences. Clin Orthop 175:147-154, 1983. [20]Keppel G: Design and Analysis: A Researcher's Handbook. Engelwood Cliffs, NJ, Prentice Hall, 1973, pp 87-88. [21]Grondin RC: The Relationship Between Biomechanical Variables, Velocity and Metabolic Rate of Normal Older Subjects Walking at Slow Speeds. Master's Dissertation. Kingston, Ontario, Canada, Queen's University, 1989. [22]Roy EA, Marteniuk RG: Mechanisms of control in motor performance: Closed loop vs motor programming control. J Exp Psychol 103:985-991, 1974. [23]Keele SW: Behavioral analysis of movement. In Brooks V (ed): Handbook of Physiology. Baltimore, MD, Williams & Wilkins, 1981, vol 2, pp 1391-1414. [24]Summers JJ: Motor programs. In Holding D (ed): Human Skills. London, England, John Wiley & Sons Ltd, 1981, pp 41-64. [25]Inglis J, Campbell D, Donald M: Electromyographic biofeedback and neuromuscular rehabilitation. Canadian Journal of Behavioral Science 8:299-323, 1976.

S Olney, PhD, is Professor, School of Rehabilitation Therapy, Queen's University, Kingston, Ontario, Canada K7L 3N6. Address correspondence to Dr Olney. GR Colborne, MSc, is Doctoral Student, Department of Anatomy, Queen's University. C Martin, BSc, is Physiotherapist in Neurological Rehabilitation, St Mary's of the Lake Hospital, 340 Union St, PO Box 3600, Kingston, Ontario, Canada K7L 5A2.