Quantitative Evaluation System of Wrist Motor Function for Stroke Patients Based on Force Feedback.

Area of Science:

• Biomedical Engineering
• Rehabilitation Medicine
• Machine Learning in Healthcare

Background:

• Traditional motor function assessment in stroke rehabilitation is subjective and lacks quantitative analysis.
• Objective evaluation of wrist motor function is crucial for personalized rehabilitation programs.
• Existing methods rely heavily on clinical experience, leading to potential inconsistencies.

Purpose of the Study:

• To develop and validate a novel quantitative evaluation system for wrist motor dysfunction in stroke patients.
• To objectively measure and refine the assessment of wrist motor function using machine learning algorithms.
• To provide a feasible tool for rehabilitation physicians to aid in clinical patient evaluation.

Main Methods:

• A force-feedback robot with embedded sensors recorded kinematic and movement data.
• Machine learning models including Random Forest (RF), Support Vector Machine Regression (SVR), K-Nearest Neighbor (KNN), and Back Propagation Neural Network (BPNN) were developed.
• The system's effectiveness was validated using data from 25 stroke patients and 10 healthy volunteers.

Main Results:

• All four machine learning models demonstrated evaluation accuracy above 88%.
• The Back Propagation Neural Network (BPNN) model achieved 94.26% accuracy.
• A high Pearson correlation coefficient (0.964) was observed between BPNN predictions and clinician scores, indicating strong agreement.

Conclusions:

• The proposed quantitative evaluation system accurately assesses wrist motor function in stroke patients.
• The BPNN model shows significant potential for precise and objective motor function evaluation.
• This system offers a feasible and refined approach to support clinical rehabilitation decision-making.