Classification of Motor Unit Action Potential Using Transfer Learning for the Diagnosis of Neuromuscular Diseases

In the assessment of neuromuscular illnesses, the motor unit action potentials (MUPs) in an electromyographic (EMG) signal are an important information source. Many methodologies in the time and frequency domains have been employed for quantitative research of EMG data since recent improvements in software EMG technology. The use of several feature extraction methods to describe MUP morphology is investigated in this article. Single classifier characteristics were used to investigate classification algorithms. To predict the class label for each MUAP, a distance weighted K-nearest neighbour (KNN) classifier was applied (Myopathic, Neuropathic, or Normal). The proposed techniques perform brilliantly in terms of overall classification accuracy, according to an exhaustive analysis of the clinical EMG database for the categorization of neuromuscular disorders.