Classifying neuromuscular diseases using artificial neural networks with applied Autoregressive and Cepstral analysis

The aim this study was to classify neuromuscular disorders using artificial neural networks (ANNs). To achieve this target, EMG signals received from normal, neuropathy, and myopathy subjects were recorded. To represent the signals adequately, matching feature parameters were obtained using Autoregressive (AR) and Cepstral analysis; executing principal component analysis was used to reduce the number of data obtained from the AR and Cepstral analysis. Following these data was used to train the ANN. Multilayer perceptron- (MLP) and radial basis function-based networks were used in the training sessions. According to our results, the combination of AR with 4-6-3 MLP topology yielded the area below the receiver operating characteristic curve of 0.954303, which is considered to be within the limits of the acceptable range.