A Hidden Semi-Markov Model based approach for rehabilitation exercise assessment

Area of Science:

• Rehabilitation Medicine
• Biomedical Engineering
• Computer Science

Background:

• Assessing patient body motion during rehabilitation is crucial for effective training.
• Current methods may lack quantitative precision or remote accessibility.
• Objective motion analysis can enhance rehabilitation outcomes.

Purpose of the Study:

• To propose and evaluate a Hidden Semi-Markov Model (HSMM) for assessing body motion in rehabilitation.
• To extract clinically relevant motion features for performance scoring.
• To provide quantitative feedback for physiotherapists and patients, especially for remote assessment.

Main Methods:

• Utilized an HSMM to analyze skeleton joint trajectories captured by RGB-D cameras.
• Extracted clinically defined exercise descriptors as motion features.
• Trained the HSMM on exemplar motion sequences.
• Validated the approach by correlating its scores with clinical assessments and Dynamic Time Warping (DTW).

Main Results:

• The HSMM-based method demonstrated strong correlation with clinical assessments.
• It showed superior discrimination between healthy and pathological movement patterns compared to DTW.
• The approach provides a reliable quantitative score for subject performance.

Conclusions:

• HSMMs offer a reliable method for quantitative assessment of motor performance in rehabilitation.
• This technology supports physiotherapists and patients with objective feedback.
• The approach is particularly suitable for remote and home-based rehabilitation monitoring.