Automatic Sorting and Analysis of Multiunit EMG Recordings

Abstract

In the clinical environment, invasive EMG recordings are currently used to acquire information about either the shape of the motor unit action potential (MUAP) waveform [3–9] or the motor unit discharge pattern [4–7]. Despite the use of selective bipolar needle electrodes, records mostly comprise MUAPs of different motor units, and, therefore, the application of a sorting procedure which separates the MUAPs of the single motor units for analysis is indicated. In the case of analyzing one single unit at a time, simple methods (like amplitude window) which can be performed on both a hardware and a software level are satisfactory, since only the extraction of a very prominent MUAP from the recorded signal is required. If, however, a simultaneous observation of several units is required for studying temporal relationships between them [5], a decomposition of the EMG signal into its constituent MUAP trains is necessary. Such a decomposition, however, represents a very complex pattern recognition task, even in the simple case of EMG recordings taken during stable weak isometric contractions; therefore, it can only be accomplished by computer software, and there are reports about efficient programs from several authors [1, 8, 10, 11]. As part of our investigations concerning motor-unit behavior in central disorders (see R. Dengler and W. Wolf, this volume) we have also developed such a decomposition program, which will be outlined by this report.