Multi-Feature, Multi-Classifier Analysis in EMG Diagnosis

In the case of difficult pattern recognition problems, the combination of the outputs of multiple classifiers using for input multiple feature sets extracted from the raw data, can improve the overall classification performance. In this work a modular neural network system in EMG diagnosis is presented where multiple features extracted from the motor unit action potential (MUAP) waveforms recorded during routine electromyographic (EMG) examination were fed into multiple classifiers, and the classification results were combined in order to improve the diagnostic yield. The feature sets computed, were: (i) the time domain parameters, (ii) the frequency domain parameters, (iii) the autoregressive coefficients, (iv) the cepstral coefficients and (v) the wavelet transform coefficients for four different wavelets (Daubechies with 4 and 20 coefficients, Chui, and Battle-Lemarie). The classifiers implemented were: (i) the back-propagation (BP), (ii) the radial basis function (RBF) network and (iii) the self-organising feature map (SOFM). The proposed system was developed for the assessment of normal subjects and subjects suffering with myopathy and motor neuron disease. It was shown that the modular neural network system enhanced the diagnostic performance of the individual classifiers making the whole system more robust and reliable.