Publications

Methods: We introduce convergence points (CP) as a marker family that describe the relation of motion between body sides across time to predict a virtual recovery point using regression techniques. The regression­based CP estimation interprets continuously recorded IMU data, i.e. gait parameters, including stride count and stride duration, which are
itself relevant for evaluating the walking abilities. The CP's, expressed in days, can help clinicians to personalise patient training, e.g. treadmill walking or balance exercises. With personalised models for estimating the physical activity (PA), we analysed differences in the affected and less­affected upper and lower body sides' movement intensity that are mapped to energy levels and expressed in metabolic equivalents. Another novel digital biomarker that can support clinicians are 3D posture cubics to quantify and visualise the functional range of motion (fROM) during therapy and free­living.

Results: In an observational clinical study, including 11 outpatients after stroke, we derived more than 620 hours of annotated movement data to investigate activities in therapy and free­living. CP estimates revealed inter-patient variability within and across gait parameters and that the patient behaviour was influenced by individual therapy schedules. The PA
analysis showed differences between the affected and less-affected body sides and the upper an lower body and that patients tend to sedentary behaviour. The fROM analysis using posture cubics confirmed that the upper arm range of motion depends on the rehabilitation context, specifically, if patients perform therapy exercises or activities of daily living.

Conclusion: Our digital biomarkers, which are specifically designed for longitudinal stroke rehabilitation, hold promise for applications in free­living.