USE OF SIGNAL REPRESENTATION TO IDENTIFY ABNORMAL MOTOR UNIT POTENTIALS IN MACRO EMG.
Macro electromyography (EMG) is a recently described recording technique allowing a nonselective recording of motor unit activity. The pick-up area of the electrode, the cannula of a modified single-fiber electrode, covers the entire motor unit territory. The motor unit potential (MUP) is obtained by averaging the cannula signals that are time locked to a single-fiber action potential. In the present report, a pattern recognition system is described which assumes no prior information about the normal and abnormal MUP's and treats the entire MUP waveform as a feature. Signal representation is used to reduce the dimensionality of the data and produced a satisfactory representation of MUP's by 20 basis vectors. The principal components of the MUP's from control subjects indicate that the features of MUP's from the control subjects form a single cluster in the feature space. Performance of the pattern recognition scheme developed was found to be superior to the conventional technique of amplitude measurements.