An algorithm for pattern recognition in needle-electromyograms - Description and clinical usefulness

Buchthal described electromyography with concentric needle electrodes, selection and classification of motor unit potentials (MUPs), and he provided normative data. Subsequently, his method was the de facto standard of quantitative electromyography. Its disadvantages are: it is time consuming, small muscles cannot be studied, and it is uncomfortable for patients. To lessen the disadvantages, computer algorithms were developed. The clinical value of these algorithms is not fully established. Methods: An own computer algorithm for selection and classification of MUPs is described. To assess its efficacy, the results of selection and classification done by the computer algorithm, by an experienced electromyographer, and by a neurology resident were compared. Subject to analysis were 75 electromyograms which were chosen at random from 10 normal subjects, from 2 patients with myopathy, and from 3 patients with a neurogenic disorder. Results: By the computer algorithm 1475, by the electromyographer 1688, and by the resident 1836 MUPs were selected. They were classified as generated by 169, 167, 195 motor units, respectively. For each motor unit, amplitude and duration of the averaged MUPs and median value of inter MUP intervals were measured automatically. The results of the computer algorithm better matched the results of the electromyographer than the results of the resident did, but differences were small and not significant. All results were similar but not equal to the values found by Buchthal. Discussion: The computer algorithm presented here helps quantitative analysis of electromyograms.