Medical decision support system for diagnosis of neuromuscular disorders using DWT and fuzzy support vector machines

The motor unit action potentials (MUAPs) in an electromyographic (EMG) signal provide a significant source of information for the assessment of neuromuscular disorders. In this work, different types of machine learning methods were used to classify EMG signals and compared in relation to their accuracy in classification of EMG signals. The models automatically classify the EMG signals into normal, neurogenic or myopathic. The best averaged performance over 10 runs of randomized cross-validation is also obtained by different classification models. Some conclusions concerning the impacts of features on the EMG signal classification were obtained through analysis of the classification techniques. The comparative analysis suggests that the fuzzy support vector machines (FSVM) modelling is superior to the other machine learning methods in at least three points: slightly higher recognition rate; insensitivity to overtraining; and consistent outputs demonstrating higher reliability. The combined model with discrete wavelet transform (DWT) and FSVM achieves the better performance for internal cross validation (External cross validation) with the area under the reciever operating characteristic (ROC) curve (AUC) and accuracy equal to 0.996 (0.970) and 97.67% (93.5%), respectively. These results show that the proposed model have the potential to obtain a reliable classification of EMG signals, and to assist the clinicians for making a correct diagnosis of neuromuscular disorders.