A morphology-based feature set for automated Amyotrophic Lateral Sclerosis diagnosis on surface electromyography

Amyotrophic Lateral Sclerosis (ALS) is a fast-progressing disease with no cure. Nowadays, needle electromyography (nEMG) the standard practice for electrodiagnosis of ALS. Surface (sEMG) emerging as more practical and less painful alternative to nEMG but still has analytical technical challenges. The objective this work was study feasibility using set morphological features extracted from sEMG support machine learning pipeline ALS diagnosis. We developed novel feature characterize based on quantitative measurements surface representation Motor Unit Action Potentials. conducted several experiments relevance proposed either individually or combined conventional sets temporal, statistical, spectral, fractal domains. validated dataset upper limb muscle data 17 patients 24 control subjects. results utility set, achieving an F1 score (81.9 ± 5.7) onset classification approach (83.6 6.9) subject approach, solely relying in right first dorsal interosseous muscle. concluded that introducing relevant automated diagnosis since it increased classifier performance during our experiments. might also help design interpretable classifiers give additional information related nature disease, being inspired by clinical interpretation sEMG.