A New Framework and Computer Program for Quantitative EMG Signal Analysis

Techniques for analyzing electromyographic signals, which estimate and detect potentials caused by active motor units in human striated muscles, are described. A framework within which these techniques are incorporated into a computer program for the quantitative analysis of EMG signals is then proposed. The resulting program allows the diagnosis of neurogenic and myogenic diseases by analyzing the waveforms of the motor unit potentials (MUP's). It also permits the research of the healthy and disturbed neuromuscular control loop by analyzing the point processes given by the activation of the single motor units. The computer program performs the decomposition process of needle electromyograms in an interactive or automatic mode. In an initial learning phase, the MUP's are estimated using a segmentation and a nearest-neighbor clustering algorithm. In the decomposition phase, a multiple decision approach with the optimal 11-error-norm and an optimization procedure are implemented which allow the separation of up to three superimposed MUP's, even when the waveforms change due to small movements of the electrodes. The decomposition results are displayed in simple diagrams to ease clinical diagnosis.