Development of an FM interferential electrical stimulator for swallowing rehabilitation

Abstract

Dysphagia (disorder of swallowing) can develop due to structural, functional, psychological, and/or medical causes. The prevalence of dysphagia in the elderly is 7%~22% per 100,000 people, and its incidence increases with age. In Japan, pneumonia is the fifth leading cause of death, with aspiration pneumonia accounting for 70% of these cases. Treatment options for dysphagia have been proposed, including transcranial direct current stimulation (tDCS), transcranial magnetic stimulation (rTMS), peripheral neuromuscular stimulation (NMES), pharyngeal electrical stimulation (PES), and drug therapy. Among other things, an electrical stimulator using peripheral neuromuscular stimulation has been developed in the USA and is the most commonly used therapy in medical practice. This device aims to improve the muscle strength required for swallowing by applying a low‐frequency electric current to the muscles. In addition, a rehabilitation device using transcutaneous electrical sensory stimulation (TESS) without muscle contraction has been developed in Japan, which can increase the frequency of swallowing without discomfort by activating the sensation in the throat with a medium frequency.In this study, a rehabilitation device will be developed using burst modulated square waves, which are high‐frequency square waves modulated by low‐frequency, devised by Hernández Arrieta. The high‐frequency component of burst‐modulated square waves can lower the impedance of the skin and induce muscle contraction with less pain than direct current or low‐frequency stimulation.In our study, we developed a multi‐point stimulation electrode for the forearm. We applied 125 patterns of electrical stimulation with 25 electrodes to five healthy subjects and were able to produce 69 different hand postures. However, because different hand postures were sometimes expressed in each experiment even when the stimulus position was the same, we sought to use machine learning models to robustly estimate the stimulus position and hand posture. The results show that the Gaussian naïve Bayes classifier has the best prediction accuracy for the thumb and index finger, where the number of bends is low, and the decision tree algorithms LightGBM and GradientBoosting classifiers for the middle, ring, and little fingers, where the number of bends is high.In the future, to use this electrical stimulator for swallowing, it is necessary to consider the shape of the electrodes to improve discomfort during stimulation, the stimulation position to stimulate appropriate muscles, and sufficient stimulation parameters to activate the swallowing muscle groups.