Intelligent Electromyograph for Early Detection of Myopathy and Neuropathy Using EMG Signals and Neural Network Model

The present work proposes developing an electromyograph to give a reliable diagnosis for detecting neuromuscular diseases. Neuropathy is a condition that affects neurons, and in myopathy, the muscle fiber does not work correctly. Developing a highly accurate diagnostic system based on EMG readings would provide a promising way to improve the evaluation of neuromuscular disorders. If the features are efficiently extracted, it is possible to obtain outstanding sorting performance. This research is carried out in two phases (hardware and software). First, the electromyogram was developed with sensors that allow the acquisition of bioelectric signals generated by the skeletal muscles with non-invasive electrodes. For recording the EMG signal, a differential pre-amplification was made, and three filters were used to obtain the minimum noise in the signal. Second, a Convolutional Neural Network (CNN) of type ResNet-34 was developed in Python. A database obtained from various articles with similar studies was built; data was a set of images of EMG signals divided into three classes: healthy, neuropathy, and myopathy. The images of these three classes are similar in time domain and frequency, so this network classifies healthy images of EMG signals from showing patterns of pathology. An EMG-based feature extraction method is proposed and implemented that uses a neural network to detect healthy conditions, myopathy, and neuropathy. Finally, according to the performance evaluation of this method, it has a precision of 98.57%.