ML-based electromyography signal analysis for assessing rehabilitation exercise execution quality.

Area of Science:

• Biomedical Engineering
• Rehabilitation Science
• Machine Learning in Healthcare

Background:

• Optimal biomechanical form is critical for successful lower limb trauma recovery through orthopedic rehabilitation exercises.
• Many patients struggle to perform rehabilitation exercises with correct form, potentially hindering recovery.
• Existing methods for assessing exercise execution quality lack real-time, objective feedback.

Purpose of the Study:

• To develop and evaluate an electromyography (EMG)-based sensor system integrated with machine learning for real-time assessment of orthopedic rehabilitation exercise quality.
• To determine the accuracy of machine learning algorithms in classifying optimal and non-optimal exercise executions.
• To explore the potential for personalized feedback to improve patient outcomes in rehabilitation.

Main Methods:

• High-density EMG data (32 electrodes) were collected from vastus lateralis and medialis muscles during four lower limb exercises (squat, hip abduction, leg raises, rocking) performed in optimal and non-optimal variations.
• A dataset of 3,040 exercise executions and 194,560 EMG recordings was generated from 19 participants.
• Support Vector Machine (SVM) and One Class SVM algorithms were employed to analyze EMG patterns and classify execution quality.

Main Results:

• A Support Vector Machine algorithm achieved an average accuracy of 83.3% (±8.8%) in classifying four distinct execution quality classes per exercise.
• A One Class SVM, trained on optimal executions, could identify unknown exercises as optimal or non-optimal with 76.6% (±5.9%) accuracy.
• The findings demonstrate the feasibility of using EMG and machine learning for objective assessment of rehabilitation exercise performance.

Conclusions:

• The proposed EMG-based sensor system with machine learning shows significant potential for evaluating exercise execution quality in real-time during rehabilitation.
• This technology can provide objective, personalized feedback to patients, potentially improving adherence and therapeutic outcomes.
• Future development could lead to advanced systems for remote monitoring and tailored rehabilitation programs, enhancing recovery after lower limb trauma.