Gaussian mixture model based clustering of Manual muscle testing grades using surface Electromyogram signals

Area of Science:

• Rehabilitation Medicine
• Biomedical Engineering
• Kinesiology

Background:

• Manual Muscle Testing (MMT) is crucial in rehabilitation but suffers from subjectivity and standardization issues.
• Electromyogram (EMG) features offer a potential objective measure for muscle strength assessment.
• Existing methods lack the precision required for accurate strength grading in clinical settings.

Purpose of the Study:

• To develop and validate an objective muscle strength grading system using EMG features.
• To classify Manual Muscle Testing (MMT) grades 4-, 4, 4+, and 5 for the Tibialis anterior muscle.
• To compare the accuracy and reliability of EMG-based grading against traditional MMT.

Main Methods:

• Fifty healthy participants underwent simultaneous MMT and EMG data acquisition for the Tibialis anterior muscle.
• Feature selection using Chi-square goodness of fit and SPEC identified key EMG parameters (Integrated EMG, RMS EMG, Waveform Length, Wilson's amplitude, Energy).
• Gaussian Mixture Model (GMM) was employed for unsupervised clustering and grade classification, with cluster evaluation via Silhouette score (0.76) and Davies Bouldin Index (0.42).

Main Results:

• The EMG-based grading system achieved a moderate agreement (Cohen's Kappa = 0.44) with manual MMT grading.
• Higher discrepancies were observed in differentiating between MMT grades 4 and 4+.
• The study demonstrated the potential for EMG-based grading to provide a more objective and standardized assessment.

Conclusions:

• EMG-based muscle strength grading offers a promising, objective alternative to subjective manual assessments.
• The proposed method can enhance the precision of strength evaluation in rehabilitation.
• This approach has the potential to be extended to various muscles and populations, aiding in personalized exercise prescription.