Machine Learning for Supporting Diagnosis of Amyotrophic Lateral Sclerosis Using Surface Electromyogram

Area of Science:

• Clinical Neurophysiology
• Biomedical Engineering

Background:

• Needle electromyogram (EMG) is a standard diagnostic tool for neuromuscular diseases.
• Amyotrophic lateral sclerosis (ALS) diagnosis often relies on invasive needle EMG.

Purpose of the Study:

• To develop and validate a noninvasive surface EMG method for supporting ALS diagnosis.
• To evaluate the diagnostic performance of specific surface EMG markers.

Main Methods:

• Characterized surface EMG patterns using clustering index, amplitude histogram kurtosis, and crossing-rate expansion kurtosis.
• Applied linear discriminant analysis with these markers to differentiate ALS patients from controls.
• Tested the method on 10 ALS patients and 11 neurologically intact subjects.

Main Results:

• The combined surface EMG markers achieved 90% diagnostic sensitivity and 100% specificity for ALS.
• This multi-marker approach outperformed single-marker analyses.
• The noninvasive method demonstrated significant diagnostic accuracy.

Conclusions:

• The proposed noninvasive surface EMG analysis is a valuable supplement to needle EMG for ALS diagnosis.
• This method offers a less invasive approach to characterizing neuromuscular function in ALS.
• High diagnostic sensitivity and specificity support its clinical utility.