Machine Learning Based Neuromuscular Disease Detection and Classification Using EMG Signal

Neuromuscular diseases (NDM) are disorders affecting the nervous system that controls the muscles. Around 15 million people are suffering from neuromuscular diseases. Electromyography (EMG) is an important technique for assessing and recording electrical activity in muscles and provides important information about muscle pain. Effective evaluation of electromyographic signals is of indisputable importance and forms the basis of neuromuscular diagnosis, developments in artificial devices, and many related applications in medicine. To improve our understanding and use of EMG signals, this study focuses specifically on EMG signals marked by the biceps muscle, precisely distinguishing individuals with neuromuscular disorders from healthy individuals. In this work, a publicly available dataset has been used where the EMG signal of patients with NDM and healthy individuals have been recorded. This work proposes a two-layer classification method that uses machine learning algorithms to detect neuromuscular disorders by extracting some specific features. In the first layer, we identify if individuals have disease or not and the second layer identifies if the individual with the disease has myopathy or neuropathy. The accuracy of the first layer is 93.3% using KNN and the second layer is 98.5% using SVM. These results not only contribute to clinical diagnosis but also encourage the expansion and improvement of this method, thus promoting the development of neuromuscular disease diagnosis using machine learning.