While the promise of wearable sensor technology to transform physical rehabilitation has been around for a number of years, the reality is that wearable sensor technology for the measurement of human movement has remained largely confined to rehabilitation research labs with limited ventures into clinical practice. The purposes of this paper are to: (1) discuss the major barriers in clinical practice and available wearable sensing technology; (2) propose benchmarks for wearable device systems that would make it feasible to implement them in clinical practice across the world and (3) evaluate a current wearable device system against the benchmarks as an example. If we can overcome the barriers and achieve the benchmarks collectively, the field of rehabilitation will move forward towards better movement interventions that produce improved function not just in the clinic or lab, but out in peoples’ homes and communities.

Background. Wearable sensors allow for direct measurement of upper limb (UL) performance in daily life. Objective. To map the trajectory of UL performance and its relationships to other factors post-stroke. Methods . Participants (n = 67) with first stroke and UL paresis were assessed at 2, 4, 6, 8, 12, 16, 20, and 24 weeks after stroke. Assessments captured UL impairment (Fugl-Meyer), capacity for activity (Action Research Arm Test), and performance of activity in daily life (accelerometer variables of use ratio and hours of paretic limb activity), along with other potential modifying factors. We modeled individual trajectories of change for each measurement level and the moderating effects on UL performance trajectories. Results. Individual trajectories were best fit with a 3-parameter logistic model, capturing the rapid growth early after stroke within the longer data collection period. Plateaus (90% of asymptote) in impairment (bootstrap mean ± SE: 32 ± 4 days post-stroke) preceded those in capacity (41 ± 4 days). Plateau in performance, as measured by the use ratio (24 ± 5 days), tended to precede plateaus in impairment and capacity. Plateau in performance, as measured by hours of paretic activity (41 ± 6 days), occurred at a similar time to that of capacity and slightly lagged impairment. Modifiers of performance trajectories were capacity, concordance, UL rehabilitation, depressive symptomatology, and cognition. Conclusions. Upper limb performance in daily life approached plateau 3 to 6 weeks post-stroke. Individuals with stroke started to achieve a stable pattern of UL use in daily life early, often before neurological impairments and functional capacity started to stabilize.

Wrist-worn accelerometry provides objective monitoring of upper-extremity functional use, such as reaching tasks, but also detects nonfunctional movements, leading to ambiguity in monitoring results. Compare machine learning algorithms with standard methods (counts ratio) to improve accuracy in detecting functional activity. Healthy controls and individuals with stroke performed unstructured tasks in a simulated community environment (Test duration = 26 ± 8 minutes) while accelerometry and video were synchronously recorded. Human annotators scored each frame of the video as being functional or nonfunctional activity, providing ground truth. Several machine learning algorithms were developed to separate functional from nonfunctional activity in the accelerometer data. We also calculated the counts ratio, which uses a thresholding scheme to calculate the duration of activity in the paretic limb normalized by the less-affected limb. The counts ratio was not significantly correlated with ground truth and had large errors (r = 0.48; P = .16; average error = 52.7%) because of high levels of nonfunctional movement in the paretic limb. Counts did not increase with increased functional movement. The best-performing intrasubject machine learning algorithm had an accuracy of 92.6% in the paretic limb of stroke patients, and the correlation with ground truth was r = 0.99 (P < .001; average error = 3.9%). The best intersubject model had an accuracy of 74.2% and a correlation of r =0.81 (P = .005; average error = 5.2%) with ground truth. In our sample, the counts ratio did not accurately reflect functional activity. Machine learning algorithms were more accurate, and future work should focus on the development of a clinical tool.

Animal models of brain recovery identify the first days after lesioning as a time of great flux in sensorimotor function and physiology. After rodent motor system lesioning, daily skill training in the less affected forelimb reduces skill acquisition in the more affected forelimb. We asked whether spontaneous human motor behaviors of the less affected upper extremity (UE) early after stroke resemble the animal training model, with the potential to suppress clinical recovery. This prospective observational study used a convenience sample of patients (n = 25, mean 4.5 ±1.8) days after stroke with a wide severity range; Controls were hospitalized for non-neurological conditions (n = 12). Outcome measures were Accelerometry, Upper-Extremity Fugl-Meyer (UEFM), Action Research Arm Test (ARAT), Shoulder Abduction/ Finger Extension Test (SAFE), NIH Stroke Scale (NIHSS). Accelerometry indicated total paretic UE movement was reduced compared to controls, primarily due to a 44% reduction of bilateral UE use. Unilateral paretic movement was unchanged. Thus, movement shifted early after stroke; bilateral use was reduced and unilateral use of the non-paretic UE was increased by 77%. Low correlations between movement time and motor performance prompted an exploratory factor analysis (EFA) revealing a 2-component solution; motor performance tests load on one component (motor performance) whereas accelerometry-derived variables load on a second orthogonal component (quantity of movement). Early after stroke, spontaneous overall UE movement is reduced, and movement shifts to unilateral use of the non-paretic UE. Two mechanisms that could influence motor recovery may already be in place 4.5 ± 1.8 days post stroke: (1) the overuse of the less affected UE, which could set the stage for learned non-use and (2) skill acquisition in the non-paretic limb that could impede recovery. Accurate UE motor assessment requires two independent constructs: motor performance and quantity of movement. These findings provide opportunities and measurement methods for studies to develop new behaviorally-based stroke recovery treatments that begin early after onset.