Design of a Data Glove for Assessment of Hand Performance Using Supervised Machine Learning

Area of Science:

• Rehabilitation Medicine
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Post-stroke recovery patients present a significant challenge to healthcare systems, increasing the workload for rehabilitation centers and physiotherapists.
• Automated assessment systems and rehabilitation robotics offer a solution to alleviate this burden, particularly for patients with a high level of recovery.

Purpose of the Study:

• To demonstrate an automated system for in-home rehabilitation assessment for post-stroke patients.
• To evaluate the system's performance using machine learning algorithms and compare different classifiers.
• To provide a means for patients to assess their recovery and transmit data to medical professionals for supervision.

Main Methods:

• Development of an automated assessment system integrating a data glove, a mobile application, and machine learning algorithms.
• Utilizing the system for performance assessment in post-stroke patients with a high recovery level.
• Comparison of two machine learning classifiers for their efficacy in assessing physical exercises.

Main Results:

• The proposed automated assessment system achieved an accuracy of 85% (±5.1%).
• Careful selection of features and classifiers contributed to the system's performance.
• The system enables remote monitoring and assessment of patient progress.

Conclusions:

• The developed automated system offers a viable solution for in-home rehabilitation assessment for post-stroke patients.
• This technology can support medical professionals by providing objective performance data, potentially improving patient outcomes and optimizing resource allocation.
• The system facilitates continuous patient monitoring and timely intervention, easing the burden on rehabilitation facilities.