AML-DECODER: Advanced Machine Learning for HD-sEMG Signal Classification-Decoding Lateral Epicondylitis in Forearm...

Area of Science:

• Biomedical Engineering
• Neuromuscular Diagnostics
• Wearable Technology

Background:

• Innovative algorithms are crucial for diagnosing and monitoring diseases like lateral epicondylitis (LE) using wearable devices.
• LE impacts various professions, causing significant daily challenges for affected individuals.

Purpose of the Study:

• To evaluate the efficacy of phase-amplitude coupling (PAC) features for diagnosing lateral epicondylitis (LE).
• To compare PAC performance against state-of-the-art and Daubechies wavelet (db4) features in predicting LE.
• To assess the potential of PAC for broader applications in neuromuscular disorder diagnostics.

Main Methods:

• High-density surface electromyography signals from forearm muscles of 14 LE patients and 14 healthy controls were analyzed.
• Phase-amplitude coupling (PAC) features were extracted and utilized to train a neural network for subject classification.
• Performance was compared against state-of-the-art and Daubechies wavelet (db4) features using specificity and sensitivity metrics.

Main Results:

• PAC features achieved 100% specificity and sensitivity in predicting unseen subjects, significantly outperforming other methods.
• PAC features demonstrated a higher Jeffries-Matusita (JM) distance, indicating robust predictive capabilities for neuromuscular diseases.
• The PAC model showed high reliability with an expected accuracy of 89% in diverse populations, validated by confidence and credible intervals.

Conclusions:

• PAC features represent a significant advancement in diagnosing lateral epicondylitis (LE).
• The findings suggest PAC's potential for developing enhanced diagnostic tools for various neuromuscular disorders.
• This research opens new avenues for understanding disease pathology and improving patient monitoring.