Machine learning-based assistance for electrode position validation in tSCS

Abstract

Spinal cord injury and multiple sclerosis can affect a patient’s walking ability and can be accompanied by spasticity. Transcutaneous spinal cord stimulation (tSCS) aims at reduction of spasticity and improvement of locomotion. This work investigates a machine learning approach to evaluate a chosen dorsal electrode position. If the position is classified as unsuitable for therapy, a recommendation for displacement is made. Classified EMG data of the posterior root muscles, evoked by a series of double-stimulation pulses with increasing intensity, and anthropometric data from 18 subjects were used to train a decision tree classifier. An average accuracy of 78% was observed.