Automated localization of nerve damage: hypothesis, technique and performance appraisal

Needle-EMG is a common procedure performed and relied on by neurologists, physiatrists and neurosurgeons to localize focal nerve damage. However, despite its popularity, there is no standardization of this diagnostic process rendering it error prone. An automated diagnostic program may help the electromyographer by providing an objective second opinion and a statistical measure of its strength. Such a program is presented and evaluated in this manuscript. The electromyographer samples a set of muscles and grades each muscle for neuropathic changes. The Program accepts the electromyographer's graded muscle set. The Program then compares it to muscle-sets derived from all possible combinations of nerve damage and finds that combination that results in the best match. The latter is the anatomical diagnosis. Eighty-five cases from the literature, each with an impartial expert diagnosis, were analyzed using this Program. The Program precisely matched the expert electromyographers' diagnoses in 79% of the cases, partially matched in 15%, and did not match in 6% of the cases. In the latter, not necessarily the Program was in error. The Program proved to be a reliable diagnostic tool that may help electromyographers by offering a second opinion. It may also assist in teaching of residents and fellows.