Myoelectric Computer Interface Training for Reducing Co-Activation and Enhancing Arm Movement in Chronic Stroke Survivors: A Randomized Trial

Abstract

Background. Abnormal muscle co-activation contributes to impairment after stroke. We developed a myoelectric computer interface (MyoCI) training paradigm to reduce abnormal co-activation. MyoCI provides intuitive feedback about muscle activation patterns, enabling decoupling of these muscles. Objective. To investigate tolerability and effects of MyoCI training of 3 muscle pairs on arm motor recovery after stroke, including effects of training dose and isometric versus movement-based training. Methods. We randomized chronic stroke survivors with moderate-to-severe arm impairment to 3 groups. Two groups tested different doses of isometric MyoCI (60 vs 90 minutes), and one group tested MyoCI without arm restraint (90 minutes), over 6 weeks. Primary outcome was arm impairment (Fugl-Meyer Assessment). Secondary outcomes included function, spasticity, and elbow range-of-motion at weeks 6 and 10. Results. Over all 32 subjects, MyoCI training of 3 muscle pairs significantly reduced impairment (Fugl-Meyer Assessment) by 3.3 ± 0.6 and 3.1 ± 0.7 (P < 10⁻⁴) at weeks 6 and 10, respectively. Each group improved significantly from baseline; no significant differences were seen between groups. Participants’ lab-based and home-based function also improved at weeks 6 and 10 (P ≤ .01). Spasticity also decreased over all subjects, and elbow range-of-motion improved. Both moderately and severely impaired patients showed significant improvement. No participants had training-related adverse events. MyoCI reduced abnormal co-activation, which appeared to transfer to reaching in the movement group. Conclusions. MyoCI is a well-tolerated, novel rehabilitation tool that enables stroke survivors to reduce abnormal co-activation. It may reduce impairment and spasticity and improve arm function, even in severely impaired patients.

Keywords

stroke rehabilitation upper extremity co-activation EMG impairment feedback function

Introduction

About 60% of stroke survivors have chronically impaired upper limb function.¹ Therapy can improve function even more than a year after stroke, but cost and availability may preclude providing long-term, intensive, therapist-based outpatient treatments to every patient. Hence, it is important to develop treatments for stroke that are not only effective but also inexpensive and portable to allow widespread use.^2,3

Impaired arm movement after stroke is caused not just by weakness, sensory loss, and spasticity but also by impaired coordination between joints^4,5 due to abnormal co-activation of muscles.^6-8 These abnormal co-activation patterns, also called abnormal synergies,^9,10 are thought to be due to impaired corticospinal tract function leading to upregulated corticobulbar pathways.¹¹ Abnormal co-activation correlates with impaired function after stroke,^5,12 and thus we hypothesized that reducing this co-activation would improve function.

We have developed a myoelectric computer interface (MyoCI) paradigm to counter this abnormal co-activation. The MyoCI maps EMGs onto cursor movements in a computer game. In a pilot study, chronic stroke survivors used MyoCI training of a single muscle pair (biceps and anterior deltoid) to reduce abnormal co-activation by 99% during the training,^13,14 but that study did not have blinded assessment of functional outcomes. This study was designed to examine the tolerability and effects on recovery of training 3 muscle pairs with MyoCI in chronic stroke survivors with moderate-to-severe impairment. It also was intended to study effects of multiple, higher doses of training and isometric versus movement-based MyoCI training on co-activation, impairment, function, spasticity, and movement kinematics. We hypothesized that training 3 muscle pairs would be tolerable and lead to reduced co-activation, impairment, and spasticity and to improved function and movement kinematics. We also hypothesized that higher dose training would be more beneficial, which has been seen in physiotherapy.¹⁵ We further hypothesized that movement-based training would lead to greater generalization of improvements in training to reduced co-activation during reaching, since motor learning transfers better to more similar contexts.¹⁶

Methods

Participants and Inclusion/Exclusion Criteria

The study protocol was approved by the Northwestern University Institutional Review Board, and each participant gave written informed consent prior to eligibility assessment. Subjects were screened by an occupational therapist to assess their affected arm’s impairment with the Fugl-Meyer upper extremity assessment (FMA-UE). We included adult, chronic stroke survivors (at least 6 months from stroke onset) who had persistent moderate to severe arm impairment (FMA-UE 8-40) and increased arm tone. We excluded subjects who (1) had impairment in vision, memory, language, or concentration; (2) received botulinum toxin on the impaired arm within 3 months; or (3) were currently participating in another research study on the arm. We performed a prospective power analysis using a repeated-measures ANOVA with a normalized effect size of 0.4 in the FMA-UE (based on preliminary data showing changes of at least 2 on FMA with an estimated standard deviation of 5¹⁷). We calculated, using 2-tailed α of .006 (Bonferroni-corrected), the power to be 0.85 given 12 subjects per group. We planned to enroll up to 42 subjects to account for attrition, estimated at 15%. The trial is registered at ClinicalTrials.gov, NCT03579992.

Subjects were simply randomized using a computer to 1 of 3 groups. Two groups used isometric activation to control the MyoCI for either 60 minutes (60-minute isometric, 60I) or 90 minutes (90-minute isometric, 90I); the third group controlled the MyoCI with unrestrained movements (90-minute movement, 90M; see below for details).

Identifying Abnormally Co-Activating Muscles

To identify abnormally co-activating muscle pairs, each subject performed a free-reaching task on weeks −2, 0, 2, 4, 6, and 10 relative to the training start. The task consisted of 3 reaches to each of 6 targets, placed at waist and shoulder height, in front of and then lateral to the impaired arm (36 reaches total, to sample a broad range of reaching movements). Subjects were instructed to bring the arm back to the rest position at the side (elbow extended maximally, shoulder adducted, not flexed nor rotated with thumb facing anteriorly) after each reach. The skin was lightly exfoliated and swabbed with alcohol to reduce impedance. Wireless, active EMG electrodes (Trigno, Delsys, Inc) were placed on the skin over the bellies of 8 muscles on the subject’s affected arm: anterior, middle, and posterior deltoids; biceps brachii, triceps (long and lateral heads); brachioradialis; and pectoralis major. Electrodes were placed according to guidelines of the Surface Electromyography for the Non-Invasive Assessment of Muscles—European Community project (seniam.org) and placement marked with a skin marker to ensure consistency across sessions. EMG was digitized at 1926 Hz during the task. A digital goniometer recorded elbow angular position at 1 kHz during reaches. The EMG envelope was obtained by high-pass filtering at 50 Hz, rectifying, then low-pass filtering at 5 Hz. Co-activation was defined as the pairwise correlation coefficient between each pair of EMG envelopes during reaching. Pairs with the largest abnormal co-activation—that is, not seen in healthy arms during reaching—were selected for MyoCI training.

Training Paradigm

The MyoCI consisted of a game developed in Python that communicated with customized C++ software to acquire the EMGs from the targeted muscle pair. The participants controlled a cursor that started in the bottom left of the screen with muscles at rest. EMG amplitudes were multiplied by a gain customized such that target distance from the starting location corresponded to 10% to 20% of maximum voluntary contraction, to equalize effort in both muscles, and were then averaged over the previous 50 ms. At each 50-ms time increment, amplitudes were mapped to orthogonal components of the cursor (0° and 90° angles), and the cursor position was determined by a vector sum of these 2 components¹³ (Figure 1B). Thus, strong co-activation of the muscle pair moved the cursor along the 45° diagonal, and only activating muscles in isolation would move the cursor along the cardinal axes. The game presented targets to the participants, who placed the cursor in a target for 0.5 seconds to successfully acquire the target. Early on, targets appeared nearer to 45° to enable subjects to acquire targets even with some co-activation. Subjects had limited time to reach the target (trial time, initially 12 seconds). If the subject entered the target but failed to hold for the duration of the hold time, the timer would reset. Failures were considered the inability to enter the target within the time limit.

Figure 1.

Participant enrollment and experimental design. (A) CONSORT recruitment flow chart. (B) Subjects controlled the MyoCI using EMGs from 2 muscles (example here showing biceps in blue and anterior deltoid in black), which were mapped to orthogonal components of cursor movement (biceps right, anterior deltoid up). (C) Examples of 2 of the game “skins” played in different weeks of training.

Between trials, subjects were given 10 seconds to relax to hold the cursor in the center target for 0.5 seconds. Some subjects had difficulty reducing their EMG activity to zero. Therefore, the software recalibrated the baseline level of EMG activation after each trial and subtracted it from the total EMG activation to control the cursor.

To help subjects learn by shaping, the game difficulty increased once a subject reached a success rate of at least 70% for 2 consecutive 10-minute runs. Difficulty was increased in 5 different parameters in the following order: target location angle (angular distance from 45°), cursor size, target size, maximum trial duration, and EMG gain. EMG gain was always maintained high enough to avoid fatigue.

Subjects trained on 3 muscle pairs for 6 sessions each over 2 weeks (Supplementary Table I). The game had 3 different “skins” to maintain motivation and remind subjects which muscles to use (Figure 1C). Subjects were pseudo-randomly assigned (without stratification) into 3 groups: 2 groups performed isometric MyoCI training for either 60 or 90 minutes per session (split into 10-minute runs with short breaks in between), and 1 group performed movement-based MyoCI training for 90 minutes per session. Subjects in the isometric groups placed their arm into a padded wrist and hand splint fixed to the armrest of a chair; the wrist and elbow were restrained using padded Velcro straps. The elbow was flexed to approximately 90° and the forearm was semi-pronated. Subjects in the non-isometric group could move their arm in any way while performing the task and were encouraged to maintain an upright posture and minimize trunk movement.

Task Performance

Task performance was measured using the number of successfully acquired targets and mean time to acquire targets, computed in each run and averaged over all runs in a session. To account for the gradually increasing difficulty, performance measures were weighted based on the difficulty level. Each weight level combined the total number of levels of each of the 5 difficulty factors (cursor size, target size, target location, maximum trial duration, and EMG gain). Subjects advanced through a mean of 51 (range of 10-80) levels of difficulty. The performance score for each run was weighted by the level of that run relative to the maximum number of levels achieved by that subject. Thus, success rate was multiplied by the weight and time was divided by the weight.

Outcome Measures

An occupational therapist blinded to the training group performed clinical assessments at −2, 0, 2, 6, and 10 weeks relative to the start of training. The therapist assessed the FMA-UE, Wolf Motor Function Test (WMFT), and Modified Ashworth Scale (MAS) on all evaluation dates. In addition, all subjects (except for one subject in the movement group) filled out the Motor Activity Log (MAL-30), a validated survey measuring amount and quality of home-based function. Distributions of WMFT times (mean over all 15 tasks) were tested for normality, and the hypothesis of normality could not be rejected, so the times (as opposed to log times) were used.¹⁸ Scores from weeks −2 and 0 evaluations were averaged together to represent baseline function before treatment. The primary outcome was change in FMA-UE scores from week 6 compared to baseline. Secondary outcome measures included the other assessments at week 6 compared to baseline, as well as all measures at week 10 compared to baseline. We also examined co-activation between the targeted muscle pairs during the free-reaching task at baseline (mean of weeks −2 and 0), weeks 2, 6, and 10.

Statistical comparisons within and among groups were performed using mixed-effects models accounting for repeated measures. Unbiased effect sizes for between-group comparisons were estimated using Hedges’ g.¹⁹ Corrections for multiple post hoc comparisons were made using the false discovery rate (FDR) method.²⁰ For one subject who did not complete week 10 evaluation, we estimated the value using the mean of weeks 2 and 6. We also performed analysis without this value and found no difference in outcomes, so we report the aggregate data with the estimated value. We compared outcomes with subjects grouped according to severity of impairment (moderate and severe for FMA ≥25 and <25, respectively).

Kinematic Analysis

We analyzed the elbow angular position to determine the change in active range of motion during the free reaching task. We determined the mean of all of the minimum angles (0° = straight) at the end of each reach (reaches were defined by peaks in anterior deltoid EMG signal). This represented the maximum elbow extension in each reach task session. We also examined the mean angle at the rest position between reaches.

Effects of Stroke Type and Location

We also analyzed the effects of stroke type (ischemic vs hemorrhagic) and location (subcortical vs subcortical + cortical) on the aforementioned outcome measures. We hypothesized that subjects with only subcortical damage would show greater improvement than those with both cortical and subcortical damage. As we only had access to MRI images in a subset of participants, and imaging reports in the remainder, we split participants into broad categories of stroke location.

Data from this study are available from the corresponding author on request.

Results

Forty-two subjects enrolled in the study; 32 (15 women, ages 27-75, mean of 6.5 years since stroke) completed the entire 6 weeks of training and were included in analysis; 31 completed the 10-week evaluation (one suffered a fractured arm unrelated to the training after week 6). Seven subjects did not meet inclusion criteria, and 2 dropped out before completing 6 weeks (Figure 1A). One subject developed nausea during training session 8 and dropped out; the other could not adhere to study appointments. All included subjects had the same therapist perform all of their evaluations; one was excluded for failing to meet this criterion. There were 12 subjects included in the 60-minute isometric group (60I), 11 in the 90-minute isometric group (90I), and 9 in the movement group (90M). Baseline characteristics of the subjects are in Table 1. Twenty-four had right hemisphere strokes and 8 had left hemisphere strokes (n = 3, 3, and 2 left hemisphere strokes in the 60I, 90I, and 90M groups, respectively). Muscle pairs trained by each subject are listed in Supplementary Table II.

Table 1.

Baseline Characteristics of 3 Subject Groups^a.

Characteristic	60I	90I	90M
Age (years)	58.5 ± 19.4	60 ± 7.2	56.8 ± 8.1
Sex (n)
Female	8	5	5
Male	4	6	5
Race (n)
White	4	5	2
Black/African American	5	6	6
Hispanic	1	1	0
Other	2	0	1
Time since stroke (years)	3.8 ± 6.2	4.3 ± 4.1	5.3 ± 3.2
FMA	17.0 ± 2.4	19.6 ± 3.0	18.7 ± 2.0
WMFT	72.4 ± 7.2	70.4 ± 9.9	70.1 ± 7.5
MAS	17.4 ± 1.9	16.1 ± 2.6	13.3 ± 1.3
MAL-Q	0.4 ± 0.1	0.6 ± 0.1	0.6 ± 0.2
MAL-A	0.5 ± 0.1	0.7 ± 0.2	0.8 ± 0.3

Abbreviations: 60I, 60-minute isometric; 90I, 90-minute isometric; 90M, 90-minute movement; FMA, Fugl-Meyer Assessment; WMFT, Wolf Motor Function Test; MAS, Modified Ashworth Scale; MAL, Motor Activity Log.

Age and time since stroke display median ± interquartile range; clinical scores display mean ± SEM.

Task Performance and Co-activation

Subjects improved in task performance over the 2 weeks spent on each muscle pair, as evinced by increased weighted success rate and reduced mean time-to-target (Figure 2A and B). This was true of all 3 subject groups.

Figure 2.

Performance and effects on co-activation. (A) Weighted success rate (mean ± SEM) improved throughout the 6 sessions trained on each muscle pair (color-coded at bottom) for all 3 color-coded subject groups: 60I (blue), 90I (red), and 90M (orange). (B) Weighted mean (±SEM) time to target decreased in each muscle pair for all 3 groups. (C) Co-activation decreased during training (solid circles; error bars: SEM) compared to reaching task at baseline (open circles). (D) Co-activation during reaching task, colored by muscle pair.

Abnormal co-activation between the targeted muscle pair decreased quickly during training for all 3 groups. There was a large decrease (54% over all subjects) within the very first session compared to co-activation during the free-reaching task (Figure 2C). Co-activation during the free-reaching task did not decrease significantly in any of the muscle pairs on average for either isometric group (Figure 2D). There were weak trends toward decreased co-activation during reaching in the first 2 muscle pairs for the movement group. While co-activation of individual muscle pairs co-activation did not significantly decrease (change of 19.4%, P = .2, paired t test from baseline to week 2 for the first muscle pair trained; change of 16.0% and 14.8%, P = .1 and .2, respectively, from baseline to weeks 6 and 10 for the second muscle pair trained), when pooled together, changes in these 2 muscle pairs at these dates were significant in the movement group (P = .015, paired t test).

Impairment Outcomes

The primary outcome measure, FMA-UE, improved significantly from baseline to week 6 in all 3 groups (Figure 3A, Table 2). While there was a trend to improved function with longer training as evinced by effect sizes (Supplementary Table III), this did not meet statistical significance at the P = .05 level. No statistical difference was seen between movement-based and isometric training in impairment. Moreover, the significantly improved gains persisted at 1 month after training stopped (Figure 3A). When all 3 groups were combined, there was a statistically significant improvement of 3.3 ± 0.6 (P < .0001) at week 6 that persisted at week 10 (3.1 ± 0.7, P < .0001). These values were close to the minimum clinically important difference (MCID) for chronic stroke, which has variably been reported as 3 for severely impaired patients²¹ and 4 to 7 for mildly to moderately impaired patients.²²

Figure 3.

Effects on impairment and function. (A) All 3 treatment groups (color-coded as in Figure 2) showed significantly (*, P < .05, FDR-corrected) reduced impairment (improved FMA-UE) at weeks 6 and 10 compared to baseline (BL). (B) Both moderately impaired (green) and severely impaired subjects (black) improved from BL. (C) WMFT scores for the treatment groups. (D) WMFT changes from baseline in moderately and severely impaired subjects. **, significantly (P < .05) greater improvement in severely impaired vs moderately impaired.

Table 2.

Mean ± SE (Derived From Mixed-Effects Models) and FDR-Corrected P Values (in Parentheses) for Each Group and Test^a.

Test	60I w6	60I w10	90I w6	90I w10	90M w6	90M w10	Mod w6	Mod w10	Sev w6	Sev w10
FMA (p-val)	3.4 ± 1.0 (.006)	3.5 ± 1.2 (.01)	3.8 ± 1.0 (.006)	2.8 ± 1.2 (.03)	2.7 ± 1.2 (.03)	2.9 ± 1.3 (.04)	5.8 ± 1.1 (.0002)	4.7 ± 1.1 (.005)	2.8 ± 0.6 (.0002)	3.0 ± 0.8 (.001)
WMFT	−4.0 ± 2.4 s (.16)	−3.3 ± 2.7 s (.24)	−7.3 ± 2.5 s (.04)	−6.8 ± 2.8 s (.07)	0.3 ± 2.8 s (.49)	−3.2 ± 2.1 s (.24)	−4.2 ± 3.6 s (.18)	0.6 ± 3.7 s (.47)	−3.9 ± 1.7 s (.045)	−6.6 ± 1.8 s (.01)
MAL-Q	0.16 ± 0.16 (0.21)	0.16 ± 0.14 (.20)	0.26 ± 0.17 (.15)	0.17 ± 0.14 (.19)	0.50 ± 0.19 (.049)	0.53 ± 0.17 (.02)	0.27 ± 0.22 (.25)	0.17 ± 0.20 (.30)	0.29 ± 0.11 (.02)	0.28 ± 0.10 (.02)
MAL-A	0.24 ± 0.15 (0.17)	0.33 ± 0.16 (.23)	0.16 ± 0.15 (.24)	0.14 ± 0.16 (.42)	0.49 ± 0.18 (.06)	0.38 ± 0.19 (.12)	0.04 ± 0.20 (.43)	−0.06 ± 0.21 (.43)	0.33 ± 0.10 (.004)	0.36 ± 0.10 (.004)
MAS	−3.9 ± 1.1 (.006)	−2.9 ± 1.6 (.09)	−3.3 ± 1.2 (.02)	−3.4 ± 1.7 (.09)	−0.8 ± 1.2 (.35)	1.0 ± 1.8 (.35)	−3.2 ± 1.6 (.08)	−2.8 ± 2.3 (.18)	−2.7 ± 0.8 (.003)	−1.8 ± 1.1 (.13)

Abbreviations: 60I, 60-minute isometric; 90I, 90-minute isometric; 90M, 90-minute movement; Mod, moderately impaired; Sev, severely impaired; FMA, Fugl-Meyer Assessment; WMFT, Wolf Motor Function Test; MAL, Motor Activity Log; MAS, Modified Ashworth Scale.

Values in bold indicate significant changes at the P = .05 level.

FMA-UE improved in both severely impaired and moderately impaired subjects (Figure 3B, Table 2). Moderately impaired subjects improved significantly more than severely impaired subjects at week 6 (P = .03) but not at week 10 (P = .15).

Functional Outcomes

We measured function using WMFT (mean time to complete each task) and MAL (Amount and Quality). Overall, subjects improved on WMFT significantly: times decreased by 3.9 ± 1.5 seconds at week 6 (P = .01) and 4.5 ± 1.6 seconds at week 10 (P = .01), substantially greater than the MCID of 1.5 seconds.²³ Groupwise WMFT changes were significant in 90I (Figure 3C, Table 2), and trending toward decrease in 60I as evidenced by the (uncorrected) 95% confidence interval (CI) [−8.9, 0.9]. Week 10 WMFT changes were nonsignificant for the 3 groups with FDR correction (Table 2). The changes trended toward decreases based on 95% CIs: [−8.8, 2.2], [−12.6, −1.1], and [−9.4, 3.2] for each of the 3 groups. There were no significant between-group differences in WMFT (P ≥ .1 in all comparisons; effect sizes in Supplementary Table II). Severely impaired subjects improved significantly at weeks 6 and 10, while moderately impaired subjects did not improve significantly (Figure 3D).

Quality of movement (MAL-Quality) improved significantly overall, though less than the MCID of 1.0²⁴: by 0.3 ± 0.10 (P = .008) at week 6 and 0.29 ± 0.09 (P = .006) at week 10. The 90M group showed significant improvement at weeks 6 and 10; the isometric groups did not significantly improve (Table 2). There were no significant between-group differences in MAL-Quality (effect sizes in Supplementary Table II). MAL-Quality improved significantly in severely impaired subjects at week 6 and 10, but did not significantly improve in moderately impaired subjects (Table 2).

Amount of movement (MAL-Amount) also improved significantly overall, though again less than the MCID: by 0.27 ± 0.09 (P = .005) at week 6 and 0.28 ± 0.10 (P = .007) at week 10. Individual groups did not improve enough to reach significance (Table 2). There were no significant between-group differences in MAL-Amount. MAL-Amount improved significantly in severely impaired subjects but not in moderately impaired subjects (Table 2).

Many subjects noticed improvements in arm function outside of MAL. One participant could open his hand much more easily. Others stated improved ability to use their impaired arm to wash dishes, bathe, get dressed, or that their arm movements were “more fluid.”

Effects on Spasticity

Over all subjects, MAS declined significantly at week 6 (−2.8 ± 0.7, P = .0002) and not quite significantly at week 10 (−2.0 ± 1.0, P = .06). Groupwise, MAS declined in both isometric groups significantly at week 6, though not significantly at week 10 (Table 2). There were no significant between-group differences in MAS (effect sizes in Supplementary Table III). Severely impaired subjects decreased significantly at week 6 but not at week 10. Moderately impaired subjects did not show significant MAS decreases (Table 2).

Effects on Elbow Movement

Unfortunately, we could only include 20 subjects in the kinematics analysis. Four subjects were excluded due to the goniometer signal saturating at angles more acute than 120°, and 8 subjects had noisy/artifactual signals in at least one session. Elbow angles during each reach improved significantly at week 6 (by 10.7 ± 4.3°, P = .03, Figure 4A black circles); improvement abated at week 10 (to 3.9 ± 3.4°, P = .19). Subjects straightened their arms more in the rest position between reaches (by 13.3 ± 4.2°, P = .004, Figure 4B) at week 6, though this also declined at week 10 to 5.2 ± 3.5° (P = .09). Due to smaller group sizes, each group showed only nonsignificant trends toward improved movement kinematics.

Figure 4.

Effects on elbow kinematics. (A) Example of elbow flexion/extension from one subject during reach task on weeks 0 (purple), 2 (blue), and 6 (green). In week 0, the subject had trouble flexing/extending, but improved in subsequent weeks. (B) Range of motion (peak-to-peak elbow flexion/extension) during reaching for the 3 groups (colors) and over all subjects (black). (C) Range of motion during rest.

Effects of Stroke Type and Location

We obtained stroke location and type information from 19 of 32 participants. Overall, the results of these analyses were mixed, and largely nonsignificant, but we report the trends given the small sample size for this analysis. Subjects with subcortical-only locations showed significantly less impairment than those with cortical and subcortical involvement (CS) at week 10 (FMA change of 3.6 ± 0.8 vs −1.7 ± 2.0, P = .04, t test) though not at week 6 (3.5 ± 0.9 vs 2.1 ± 0.6, P = .25). WMFT showed a nonsignificant trend toward more improvement in subjects with CS than with subcortical-only, while MAL-Q and MAS showed nonsignificant trends in the opposite direction; MAL-A did not show any trend. There was no clear trend for hemorrhagic versus ischemic stroke subtypes in any outcome at week 6, while at week 10, there was a small, nonsignificant trend in favor of ischemic strokes in WMFT and MAS scores.

Discussion

Myoelectric computer interface training enables chronic stroke survivors to effectively decouple abnormally co-activating muscles. MyoCI training of 3 muscle pairs was well-tolerated at all doses; only one subject dropped out—from nausea that was unlikely to be due to training, given that it occurred in the eighth session. MyoCI training reduced arm impairment significantly in all groups, approximately at the MCID of FMA for severely impaired survivors. Functional improvement was also exhibited over all subjects combined. Although individual group changes in WMFT were not significant, there were trends toward decreases in each group. MAS scores decreased noticeably in the isometric groups at 6 weeks, though not in the movement group. Elbow active range of motion also increased at 6 weeks, both during reaching and in the rest position. There were no statistically significant between-group differences in FMA, WMFT, MAL, or MAS.

While many approaches to motor rehabilitation after stroke have been developed, very few have targeted co-activation. EMG biofeedback of single muscles had mixed results,^25,26 and largely sought to strengthen muscles or reduce spasticity—not co-activation—by “downtraining” the spastic muscle using simple auditory or visual feedback. One study using biofeedback with the goal of increasing triceps activity did not show significant improvement in elbow AROM.²⁷ One small study (2 stroke survivors) provided EMG feedback about wrist flexors and extensors while moving a wrist manipulandum for 2 months of training.²⁸ This study showed slightly reduced duration of co-activation during training, and improved movement, in one patient, but did not test any outcomes of function or impairment.²⁸ Few EMG biofeedback studies have measured outcomes using FMA-UE. One, a randomized trial of EMG biofeedback plus physiotherapy versus physiotherapy alone, in the first 8 weeks poststroke, did not show significantly greater improvement in FMA with biofeedback.²⁹ Another randomized controlled trial of 80 patients (mixed acute and chronic phase) showed significantly more improvement on wrist and hand portions of FMA (gain of 4.5 vs 1.2 points) with wrist EMG biofeedback than with conventional physiotherapy.³⁰ However, the most recent Cochrane review of EMG biofeedback concluded that there was no treatment benefit when combining all studies, and positive trials were small and generally poorly designed.³¹

One paradigm has addressed abnormal joint torque couplings using visual feedback of joint torques.¹⁴ An alternative paradigm used a robot to provide progressive loading of shoulder abduction to reduce abnormal joint couplings.³² Those paradigms are not well-suited for use outside the lab or academic medical center, due to the size and expense of the robotic equipment and need for specialized assistance. Furthermore, there is a redundant mapping between muscles and joint torques, such that subjects attempting to decouple joint torques may be recruiting extra muscles to achieve the desired torque output. This recruitment of extra muscles may change depending on the task, and the learned behavior may not transfer to new tasks. In contrast, the MyoCI enabled users to target specific muscles to decouple. Indeed, subjects quickly learned to decouple abnormally co-activating muscle pairs during training. While this decoupling did not transfer to the reaching task for the isometric groups, the movement group did show some decoupling of the targeted muscles during the reaching task. This suggests that there was better generalization of the motor learning in the movement group than the isometric groups.

The results presented here strongly suggest that even severely impaired patients can benefit from MyoCI training (eg, Figure 3B and D, Table 2). These patients have the most abnormal co-activation⁷ and are the most in need of new therapies³³ since conventional techniques provide little benefit when patients lack sufficient movement to participate.³⁴ MyoCI therapy removes this requirement and could at minimum enable severely impaired survivors to benefit from other therapies.

It is encouraging that the participants overall exhibited functional improvement with the therapy, both inside and outside of the lab. Moreover, this functional improvement was substantially greater than the MCID of WMFT in chronic stroke subjects. This suggests that MyoCI training could improve arm function by a clinically meaningful amount. Notably, participants were able to extend their elbows more, both at rest and during reaching, at the end of training. This is significant in that abnormal co-activation is more prominent with increased loading of the shoulder.³⁵

There are some limitations to this study. It was a relatively small sample, and we did not meet our goal of 12 subjects in each group. We did not find statistically significant differences between experimental groups in any outcome. Effect sizes for the primary outcome (FMA) were smaller than we anticipated in the power analysis, which appeared to be due to smaller differences among the group means than we expected. Also, there was no active control, so it is possible that improvement was due to patients merely using their affected arm more than usual, counteracting learned nonuse. Our ongoing follow-up study includes a sham control for this reason.

There were no effects of stroke type, and mixed effects of stroke location, on outcomes. There were nonsignificant trends toward improved outcomes with subcortical only, as opposed to subcortical plus cortical, strokes. Unfortunately, we were limited in our ability to further characterize lesion location, and these results (including the lack of significant differences in most outcomes due to location) could have been affected by the limited number of imaging results available to us. It is possible that these effects could be due to larger size of cortical plus subcortical strokes compared with subcortical-only strokes; our data did not permit this to be assessed. We plan to examine these questions more closely in the follow-up study.

One perhaps surprising result of this study was the reduction of MAS scores. Abnormal co-activation is a disorder of motor control and putatively caused by a different mechanism—the use of corticobulbar pathways, in particular those to reticulospinal tracts³⁶—than spasticity. It is unlikely that changes in spasticity would explain the changes in function that we see here, especially since spasticity is only seen during reaching when the arm is supported,³⁷ which is not the case here. While MAS is controversial in its ability to accurately measure spasticity,³⁸ it is still possible that the MyoCI training did have effects at the spinal cord level, which could have affected spasticity as well as co-activation. Indeed, abnormal co-activation can modulate spasticity,³⁹ so reducing co-activation could also reduce spasticity. Further investigation will be needed to determine the location of motor learning with MyoCI training.

MyoCI offers several potential advantages. It can be made inexpensively, which could enable widespread use, in contrast to more expensive therapies such as robots. Furthermore, we are investigating a wearable version that enables home-based training, allowing higher training intensity, something crucial to improvement.⁴⁰ Integration of games into training made it enjoyable; sustaining motivation is another critical factor to recovery.⁴¹ A wearable, inexpensive device, and enjoyable training together could enable patients to continue training for longer durations or with “tune-up” sessions. Thus, even if the changes from training end up not persisting long-term, MyoCI training has potential to benefit patients. Finally, since it addresses a different mechanism of stroke impairment than traditional therapies, it could benefit stroke survivors to combine MyoCI with other stroke therapies.

Conclusion

This study provides results suggesting that MyoCI training may improve function and reduce impairment even years after stroke. MyoCI training appears safe and well-tolerated. Since it could be made wearable, it represents a promising new treatment for stroke.

Supplemental Material

MyCI_paper_v2_Supp_Material – Supplemental material for Myoelectric Computer Interface Training for Reducing Co-Activation and Enhancing Arm Movement in Chronic Stroke Survivors: A Randomized Trial

Supplemental material, MyCI_paper_v2_Supp_Material for Myoelectric Computer Interface Training for Reducing Co-Activation and Enhancing Arm Movement in Chronic Stroke Survivors: A Randomized Trial by Emily M. Mugler, Goran Tomic, Aparna Singh, Saad Hameed, Eric W. Lindberg, Jon Gaide, Murad Alqadi, Elizabeth Robinson, Katherine Dalzotto, Camila Limoli, Tyler Jacobson, Jungwha Lee and Marc W. Slutzky in Neurorehabilitation and Neural Repair

Footnotes

Acknowledgements

We thank Zachary Wright for assistance with the MyoCI software, and Kathy Kopka for assistance with outcome evaluations.

Supplementary material for this article is available on the Neurorehabilitation & Neural Repair website at along with the online version of the article.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported in part by National Institutes of Health Grants R21NS084069, R01NS099210, and UL1TR001422.

ORCID iDs

Tyler Jacobson

Marc W. Slutzky

References

Wade

Langton-Hewer

Wood

Skilbeck

Ismail

The hemiplegic arm after stroke: measurement and recovery. J Neurol Neurosurg Psychiatry. 1983;46:521-524.

Page

Sisto

Levine

McGrath

RE.

Efficacy of modified constraint-induced movement therapy in chronic stroke: a single-blinded randomized controlled trial. Arch Phys Med Rehabil. 2004;85:14-18.

Dobkin

BH.

Clinical practice. Rehabilitation after stroke. N Engl J Med. 2005;352:1677-1684.

Beer

Dewald

Rymer

WZ.

Deficits in the coordination of multijoint arm movements in patients with hemiparesis: evidence for disturbed control of limb dynamics. Exp Brain Res. 2000;131:305-319.

Zackowski

Dromerick

Sahrmann

Thach

Bastian

AJ.

How do strength, sensation, spasticity and joint individuation relate to the reaching deficits of people with chronic hemiparesis?

Brain. 2004;127(pt 5):1035-1046.

Dewald

Pope

Given

Buchanan

Rymer

WZ.

Abnormal muscle coactivation patterns during isometric torque generation at the elbow and shoulder in hemiparetic subjects. Brain. 1995;118(pt 2):495-510.

Roh

Rymer

Perreault

Yoo

Beer

RF.

Alterations in upper limb muscle synergy structure in chronic stroke survivors. J Neurophysiol. 2013;109:768-781.

Cheung

Turolla

Agostini

et al . Muscle synergy patterns as physiological markers of motor cortical damage. Proc Natl Acad Sci U S A. 2012;109:14652-14656.

Twitchell

TE.

The restoration of motor function following hemiplegia in man. Brain. 1951;74:443-480.

10.

Brunnström

Movement Therapy in Hemiplegia: A Neurophysiological Approach. New York, NY: Harper & Row; 1970.

11.

McMorland

Runnalls

Byblow

WD.

A neuroanatomical framework for upper limb synergies after stroke. Front Hum Neurosci. 2015;9:82.

12.

Chae

Yang

Park

Labatia

Muscle weakness and cocontraction in upper limb hemiparesis: relationship to motor impairment and physical disability. Neurorehabil Neural Repair. 2002;16:241-248.

13.

Wright

Rymer

Slutzky

MW.

Reducing abnormal muscle coactivation after stroke using a myoelectric-computer interface: a pilot study. Neurorehabil Neural Repair. 2014;28:443-451.

14.

Ellis

Holubar

Acosta

Beer

Dewald

JPA

. Modifiability of abnormal isometric elbow and shoulder joint torque coupling after stroke. Muscle Nerve. 2005;32:170-178.

15.

Pollock

Baer

Campbell

et al . Physical rehabilitation approaches for the recovery of function and mobility following stroke. Cochrane Database Syst Rev. 2014;(4):CD001920.

16.

Krakauer

JW.

Motor learning: Its relevance to stroke recovery and neurorehabilitation. Curr Opin Neurol. 2006;19:84-90.

17.

Guarino

Krebs

et al . Multicenter randomized trial of robot-assisted rehabilitation for chronic stroke: methods and entry characteristics for VA ROBOTICS. Neurorehabil Neural Repair. 2009;23:775-783.

18.

Wolf

Catlin

Ellis

Archer

Morgan

Piacentino

Assessing Wolf Motor Function Test as outcome measure for research in patients after stroke. Stroke. 2001;32:1635-1639.

19.

Hedges

LV.

Distribution theory for Glass’s estimator of effect size and related estimators. J Educ Stat. 1981;6:107-128.

20.

Benjamini

Hochberg

Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B Stat Methodol. 1995;57:289-300.

21.

Guarino

Richards

et al . Robot-assisted therapy for long-term upper-limb impairment after stroke. N Engl J Med. 2010;362:1772-1783.

22.

Page

Fulk

Boyne

Clinically important differences for the upper-extremity Fugl-Meyer scale in people with minimal to moderate impairment due to chronic stroke. Phys Ther. 2012;92:791-798.

23.

Lin

Hsieh

Chen

Jang

Liu

JS.

Minimal detectable change and clinically important difference of the Wolf Motor Function Test in stroke patients. Neurorehabil Neural Repair. 2009;23:429-434.

24.

Simpson

Eng

JJ.

Functional recovery following stroke: capturing changes in upper-extremity function. Neurorehabil Neural Repair. 2013;27:240-250.

25.

Schleenbaker

Mainous

3rd . Electromyographic biofeedback for neuromuscular reeducation in the hemiplegic stroke patient: a meta-analysis. Arch Phys Med Rehabil. 1993;74:1301.

26.

Wolf

SL.

Electromyographic biofeedback applications to stroke patients. A critical review. Phys Ther. 1983;63:1448-1459.

27.

Wolf

Catlin

Blanton

Edelman

Lehrer

Schroeder

Overcoming limitations in elbow movement in the presence of antagonist hyperactivity. Phys Ther. 1994;74:826-835.

28.

Davis

Lee

RG.

EMG biofeedback in patients with motor disorders: an aid for coordinating activity in antagonistic muscle groups. Can J Neurol Sci. 1980;7:199-206.

29.

Crow

Lincoln

Nouri

De Weerdt

The effectiveness of EMG biofeedback in the treatment of arm function after stroke. Int Disabil Stud. 1989;11:155-160.

30.

Doğan-Aslan

Nakipoğlu-Yüzer

Doğan

Karabay

Özgirgin

The effect of electromyographic biofeedback treatment in improving upper extremity functioning of patients with hemiplegic stroke. J Stroke Cerebrovasc Dis. 2012;21:187-192.

31.

Woodford

Price

EMG biofeedback for the recovery of motor function after stroke. Cochrane Database Syst Rev. 2007;(2):CD004585.

32.

Ellis

Carmona

Drogos

Dewald

JPA

. Progressive abduction loading therapy with horizontal-plane viscous resistance targeting weakness and flexion synergy to treat upper limb function in chronic hemiparetic stroke: a randomized clinical trial. Front Neurol. 2018;9:71.

33.

Buch

Shanechi

Fourkas

Weber

Birbaumer

Cohen

LG.

Parietofrontal integrity determines neural modulation associated with grasping imagery after stroke. Brain. 2012;135(pt 2):596-614.

34.

Nowak

Grefkes

Ameli

Fink

GR.

Interhemi-spheric competition after stroke: Brain stimulation to enhance recovery of function of the affected hand. Neurorehabil Neural Repair. 2009;23:641-656.

35.

Beer

Dewald

JPA

Dawson

Rymer

WZ.

Target-dependent differences between free and constrained arm movements in chronic hemiparesis. Exp Brain Res. 2004;156:458-470.

36.

McPherson

Chen

Ellis

Yao

Heckman

Dewald

JPA

. Progressive recruitment of contralesional cortico-reticulospinal pathways drives motor impairment post stroke. J Physiol. 2018;596:1211-1225.

37.

Ellis

Schut

Dewald

JPA

. Flexion synergy overshadows flexor spasticity during reaching in chronic moderate to severe hemiparetic stroke. Clin Neurophysiol. 2017;128:1308-1314.

38.

Alibiglou

Rymer

Harvey

Mirbagheri

MM.

The relation between Ashworth scores and neuromechanical measurements of spasticity following stroke. J Neuroeng Rehabil. 2008;5:18.

39.

McPherson

Stienen

Drogos

Dewald

JP.

Modification of spastic stretch reflexes at the elbow by flexion synergy expression in individuals with chronic hemiparetic stroke. Arch Phys Med Rehabil. 2018;99:491-500.

40.

Kwakkel

Wagenaar

Koelman

Lankhorst

Koetsier

JC.

Effects of intensity of rehabilitation after stroke a research synthesis. Stroke. 1997;28:1550-1556.

41.

Robertson

Ridgeway

Greenfield

Parr

Motor recovery after stroke depends on intact sustained attention: a 2-year follow-up study. Neuropsychology. 1997;11:290-295.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.02 MB