Classification of Normal, Neuropathic, and Myopathic Electromyograph Signals Using Nonlinear Dynamics Method

Muscles are the vital organ responsible for the movement of the body. Neuropathies and myopathies are diseases associated with muscles. In this work, we use Electromyograph (EMG) signals are taken from the dorsiflexion of the foot to perform an automatic classification of subjects with and without neuro-muscular disorders. EMG signals are non-linear and non-stationary in nature. Hence, it is very difficult to analyse these signals using conventional statistical and frequency domain methods. Hence, we have used non-linear dynamics methods to extract the hidden information in the EMG signals. These signals are analysed by extracting five features of Approximate Entropy (ApEn), Correlation Dimension (CD), Hurst Exponent (H), Fractal Dimension (FD) and Largest Lyapunov Exponent (LLE). Afterthe statistical analysis, these selected non-linear features were fed to four classifiers namely Decision Tree (DcT), Fuzzy, K-Nearest Neighbour (KNN) and Probabilistic Neural Network (PNN) classifiers to select the best classifier. The performance of these classifiers were evaluated using accuracy, sensitivity, specificity, and positive predictive value. We observed that the KNN classifier presented the highest accuracy of 99.3%, sensitivity of 99.6% and specificity of 100%, while the other classifiers were very closely behind. Thus, the proposed automated technique has the ability to clearly differentiate between normal and diseased patients' EMG signals (myopathy and neuropathy).