Detection of Myopathy and ALS Electromyograms Employing Modified Window Stockwell Transform

Abstract

In this article, a novel technique for detection and classification of electromyograms is proposed employing the modified window Stockwell transform. Instead of using a conventional Gaussian window, a modified signal-dependent adaptive Gaussian window is proposed for improved analysis of electromyograms in joint-time frequency frame. The parameters of the proposed modified Gaussian window are optimized using a particle swarm optimization algorithm to maximize the energy concentration measure in time-frequency plane. The electromyograms of myopathy and amyotrophic lateral sclerosis disorders are subsequently probed using the proposed modified Gaussian window to obtain their respective time–frequency representations. From the transformed signals in the joint-time frequency domain, several new features are proposed, and student's t -test is conducted to examine their statistical significance. Using the selected features, classification of myopathy and amyotrophic lateral sclerosis disorders is done using four benchmark classifiers. Investigations reveal that the highest mean classification accuracy of 98.58% is achieved in this article, which proves the efficacy of the proposed method for automated diagnosis of neuromuscular disorders.