Sensing System for Plegic or Paretic Hands Self-Training Motivation

Area of Science:

• Neuroscience and Rehabilitation Engineering
• Human-Computer Interaction
• Medical Technology

Background:

• Stroke survivors with paretic or plegic hands need consistent exercises for neuroplasticity and joint mobilization.
• Existing hand exercise devices often require some patient control or offer limited engagement.
• Cognitive impairments post-stroke, like forgetfulness, hinder consistent self-rehabilitation efforts.

Purpose of the Study:

• To present Przypominajka v2, a system designed to support, remind, and motivate stroke patients during self-exercising.
• To integrate a glove-based device with on-device machine learning for exercise scoring.
• To develop a comprehensive system including a tablet interface and a web application for therapists.

Main Methods:

• Development of a glove-based sensor system with on-device machine learning for exercise classification.
• Evaluation of on-device inference feasibility and exercise classification accuracy in healthy participants.
• A case study involving a stroke patient with a paretic hand to assess system usability and effectiveness.

Main Results:

• On-device anomaly classification achieved 91.3% accuracy and 91.6% F1 score, with lower performance for new users (78% and 81%).
• The case study indicated a positive patient response to using Przypominajka for hand exercises.
• Identified usability issues with the sensor glove, specifically regarding ease of donning and clarity of instructions.

Conclusions:

• Sensor systems with on-device machine learning can effectively support stroke patient rehabilitation by scoring exercises and enhancing motivation.
• Przypominajka v2 demonstrates a novel approach to aiding self-rehabilitation for post-stroke hand paresis.
• Further refinement of the sensor glove interface is recommended to improve user experience and adherence.