Feature extraction from multifractal spectrum of electromyograms for diagnosis of neuromuscular disorders

Abstract

In this contribution, a novel technique for discrimination of myopathy, amyotrophic lateral sclerosis (ALS) and healthy electromyograms is proposed using multifractal detrended fluctuation analysis (DFA). Electromyograms are aperiodic and nonstationary electrical signals which represent the complex dynamics of skeletal muscle tissues and nerve cells activities within the human body. In this study, non-linear and dynamic fluctuations of noisy and chaotic electromyograms are analysed using fractal geometry. Electromyography (EMG) signals of myopathy, ALS and healthy disorders were collected from an online existing database and the non-linear dynamics were initially characterised using multifractal DFA. Following this, five novel feature parameters were extracted from the multifractal spectrum of respective EMG signals. Analysis of variance test was conducted on the selected features to examine their statistical significance. Finally, classification of myopathy, ALS and healthy electromyograms was done using support vector machine and k-nearest neighbour classifiers. In this study, four classification tasks are reported and it was observed that the performance of the proposed method is reasonably satisfactory in discriminating different categories of electromyograms. In addition, the proposed method is also found to deliver comparable and even better results in comparison with the existing methods, studied on the same data set.