Preliminary Examination of the Ability of a New Wearable Device to Capture Functional Hand Activity After Stroke
A reliable measure of movement repetitions is required to assist in determining the optimal dose for maximizing upper limb recovery after stroke. This study investigated the ability of a new wearable device to capture reach-to-grasp repetitions in individuals with stroke.
Methods—Eight individuals with stroke wore an instrumented wrist bracelet while completing 12 upper limb activities. Participants completed 5 and 10 repetitions of each activity on 2 separate sessions (time 1 and time 2) and completed clinical assessments (Fugl-Meyer Upper Extremity Assessment and Action Research Arm Test). Mean reach-to-grasp counts (ie, hand counts) were compared across activities. Scaling properties were assessed by the ratio of 10 repetitions to 5 repetitions for the activities (ie, expected value of 2). Bland-Altman diagrams were used to examine agreement between time 1 and time 2 counts.
Results—The wrist bracelet averaged 0 to 0.6 hand counts per repetition for the arm-only and hand-only activities and averaged 1 to 2 counts per repetition of the reach-to-grasp activities. The mean ratio of 10 repetition to 5 repetition counts was ≈2 for all of the reach-to-grasp activities. Mean differences from time 1 to time 2 wereConclusions—
These preliminary results provide evidence that the wrist bracelet is able to capture hand counts over a variety of tasks in a consistent manner. This wrist bracelet could be further developed as a tool to record dose of upper limb practice for research or clinical practice, as well as providing motivation and accountability to patients participating in treatments requiring upper limb movement repetitions. Currently, there are limitations in interpreting the impact of impairment and common compensatory movements on hand counts, and it would be valuable for future studies to explore these effects.
Animal studies suggest thousands of challenging reach-to-grasp movements are necessary to drive functional recovery after stroke.1 Current doses of movement practice during stroke rehabilitation do not come close to those obtained in animal studies2; however, the optimal dose of therapy for humans has remained elusive. Conflicting evidence has resulted in unanswered questions regarding the optimal number and frequency of repetitions of the affected upper limb inside and outside of therapy.3 A reliable measure of movement repetitions is required to begin unpacking the effect of these variables. Counting of reach-to-grasp movements using human observation is laborious and subjective, and current wearable sensors, which rely on accelerometers, do not provide clinically interpretable information such as number of reaches or grasps. Although an instrumented glove can provide accurate characterization of hand movement, we propose to develop a unique device in individuals with stroke that will permit monitoring of functional reach-to-grasp activities through a wrist band that does not interfere with sensory or motor functions of the hand.
Methods
Anonymized data will be shared on request from any qualified investigator. Additional Methods information is available in the online-only Data Supplement.
Population
Community-dwelling individuals were recruited from a volunteer database if they had residual right-sided upper limb impairment resulting from a stroke and were able to reach and grasp a variety of items (ie, cup and plate).
Device
A prototype version of the TENZR wrist band developed by BioInteractive Technologies was used (Figure [A]). The wrist band can track arm movements, wrist movements, and movement of the tendons of the wrist using proprietary sensing technologies and includes inertial measurement units, proximity sensors, and force myography sensors. Force myography sensors measure the localized pressure exerted by the wrist musculoskeletal complex on the band. The device outputs a unit called hand counts, which returns a count of one when the band senses arm movement before or after detecting both wrist activity (ie, relative wrist flexion or extension) and hand activity (ie, a change in the force myography signal pattern). This was to approximate a functional hand movement that involves the initial interaction with an object (ie, reach and grasp) or the final interaction with an object (when the hand releases and the arm moves away from the object). Please see https://www.ahajournals.org/journal/str for more description of the algorithm. A slight pause in between subsequent reach and grasp movements is required (ie, 0.1–0.2 seconds) for the band to detect the start of a new distinct reach-to-grasp task. Finally, the β-device used for testing was only configured for the right arm.
Procedures
Participants completed 12 activities while wearing the band on their right wrist. The activities were selected to represent activities that only involved the arm or hand and those that represented a range of functional reach-to-grasp activities (Table 1). Please see https://www.ahajournals.org/journal/str for the explanation of the expected hand counts per task. Participants performed each activity 5 and then 10×. This protocol was completed twice (time 1 and time 2) separated by a period of time where the device was removed (at least 30 minutes). Participants completed clinical assessments and rested during the interim period of time when the device was not worn. The Fugl-Meyer Upper Extremity Assessment4 and Action Research Arm Test5 were administered to characterize the upper limb impairment and functional level of the sample and are recommended by the Stroke Recovery and Rehabilitation Roundtable for capturing sensorimotor recovery of the upper limb after stroke.6 Ethics approval for this project was obtained from local university and hospital research ethics boards. Written informed consent was obtained from all participants.
Statistical Analysis
The mean and CI of the scaling ratio (hand counts10 reps/hand counts5 reps), with an expected value of 2, was calculated to examine the effect of doubling the activity repetitions on the relationship between counts. Agreement between time 1 and time 2 counts was examined using Bland-Altman plots for all activities.
Results
The participants were 8 people with right-sided weakness due to a stroke with mild-to-moderate impairment as captured by the clinical scales (Table 2). Table 1 shows the expected and observed hand counts for the 12 activities. The wrist band averaged 0 to 0.6 counts per repetition for the activities that involved just the hand or arm (expected zero) and ranged from 0.9 to 2.0 counts per repetition for the reach-to-grasp activities (Table 1). Counts for the arm activities were low but higher than expected and varied more between participants than counts for the hand activities (Table 1). The ratio of counts in response to doubling activity repetitions was ≈2 for all 5 reach-to-grasp activities. The mean difference between time 1 and time 2 approached zero (Figure [B]). The one exception was grasping a plate (mean difference/repetition, 0.7; Figure [B]), which also had larger limits of agreement. The participants with the most impairment (participants 4 and 5), as well as a participant with the least impairment (participant 2), had differences close to the limits of agreement for 2 of 5 of the reach-to-grasp activities (Figure [B]).
Discussion
Our study provides preliminary evidence that an instrumented wrist band can capture functional reach-to-grasp repetitions among individuals with mild-to-moderate impairment following stroke. This is the first study to examine the utility of a wrist band that incorporates force myography to capture functional hand activity following stroke. This study represents advancement on previous work in a highly controlled environment that found that force myography placed at the forearm (unlike the wrist in this study) of individuals with stroke, combined with machine learning algorithms, was able to detect grasping of an object with high accuracy.7
The device tested in our study was able to differentiate between activities that only involved the hand or arm and functional activities that involved reaching and grasping objects. In addition, the hand counts increased by a factor of 2 as expected when the number of repetitions were doubled. Further evidence of measurement consistency was provided when mean differences between test-retest administrations of the protocol approached zero for the majority of the activities. In addition, the observed hand counts matched our expectations for the majority of activities. We did not expect to observe hand counts for the arm activities. However, we could not control the exact motions or compensations that occurred, and some individuals may have extended the wrist while clearing the divider of the box and blocks or closed their hand during an arm swing while walking, which would have triggered a hand count. As a result, compensatory movements would potentially introduce more intersubject hand count variability for the same task. Future work that incorporates measures of quality of movement could further examine the effect of common compensatory movements after stroke on hand counts. Moreover, future studies should also explore the utility of this device to capture daily activities in a nonlaboratory setting.
Some of the observed variation in counts is due to the natural variation in performing the tasks. For example, some individuals used little wrist motion during the Box and Block task, and the wooden block was so small (2.5 cm) and light that little force was required. Consequently, the low wrist and hand activity may not have triggered a count. In contrast, some individuals may have generated close to 2 counts per repetition if they used a sufficient amount of wrist and hand activity to surpass the threshold and if they chose to pause on one side of the Box and Block task, which would then trigger a second count.
One might have expected poorer reliability with more impaired participants as greater motor pattern variability is a defining feature of upper limb movement following stroke. While the 2 participants with the greatest level of impairment were closer to the limits of agreement in 2 of 5 activities, so too was the participant with the least impairment; future studies with larger samples of people across the mild-to-moderate impairment spectrum are required to further explore the effect of impairment on hand counts. Finally, higher hand counts were obtained for participants who were noted to have hit their wrist or the band on a surface during testing (Figure [B]: box and blocks, participant 4; open jar, participant 5). Future versions of this device are expected to remove hand counts due to bumping. Furthermore, these preliminary results are only generalizable to individuals with mild-to-moderate impairment. A variant of the wrist band that customizes the wrist and hand activity thresholds for the user may be more suitable to individuals with severe impairment and would need to be investigated in future studies.
Conclusions
This study suggests that a wrist-worn device is able to consistently capture functional reach-to-grasp repetitions in individuals with mild-to-moderate upper limb stroke impairment. This wrist bracelet could be further developed as a tool to record the dose of upper limb practice for research or clinical practice, as well as provide motivation and accountability to patients participating in treatments requiring upper extremity movement repetitions.
This study was supported by the Canadian Institute of Health Research, Natural Sciences and Engineering Research Council, and Canada Research Chair Programs.
Dr Menon has a vested interest in commercializing the technology and may benefit financially from its potential commercialization through BioInteractive Technologies. The other authors report no conflicts.
Footnotes
The online-only Data Supplement is available with this article at https://www.ahajournals.org/doi/suppl/10.1161/STROKEAHA.119.026921.
Customer comments
No comments were found for Preliminary Examination of the Ability of a New Wearable Device to Capture Functional Hand Activity After Stroke. Be the first to comment!