1000 Section - Remote Monitoring of Upper Limb Home Exercise Performance in Post-Stroke Patients. A Pilot Study

Remote monitoring of home exercise performance after stroke is a pressing issue in modern rehabilitation. Human Pose Estimation (HPE) is considered a promising non-invasive approach capable of analyzing skeletal kinematics in real time.

The aim of this study was to evaluate the feasibility of using a markerless HPE approach for remotely monitoring home exercise performance in patients after stroke over a long period.

Design:

The study included six patients with hemiparesis who performed home exercises for 6–9 months. Two exercises were analyzed: "hand to mouth" and "180° arm abduction." During the exercises, the patient received auditory and visual cues based on performance errors.

Results:

For the "180° arm abduction" exercise, 762 sessions were analyzed. The most common error was elbow flexion, occurring in 94.7% of sessions. Shoulder elevation errors occurred in 51.6% of sessions, trunk tilt errors in 38.6%, and head tilt errors in 24.4%. Some patients showed covariance in the frequency of "shoulder elevation" and "arm flexion" errors, which may reflect the interrelated restructuring of compensatory strategies during training.

For the hand-to-mouth exercise, 700 sessions were analyzed. The most common error was "shoulder elevation," occurring in 64.6% of sessions; head tilt errors occurred in 21.2%; and trunk tilt errors occurred in 15.1%. Some patients demonstrated persistent combinations of shoulder elevation, trunk tilt, and head tilt errors.

Conclusion: Long-term monitoring of HPE in home rehabilitation allows for a quantitative description of the profile of compensatory motor strategies and the dynamics of their changes. In the 180° abduction exercise, the predominant pattern is associated with elbow flexion, while in the hand-to-mouth exercise, shoulder elevation is more frequently recorded. Some patients demonstrated consistent combinations and coordinated dynamics of individual errors. In the future, this approach may be used as a tool for refining rehabilitation prognosis and developing personalized training programs.