Combination of motion data and electromyography for threshold-based swallow onset detection

Abstract

Dysphagia – difficulty in swallowing – represents a restriction for patients. Approaches like biofeedback-based training and functional electrical stimulation (FES) arise as supporting methods. Both require online swallow onset detection for triggering, whereby non-invasive measurement methods should be preferred. In this pilot work, a threshold-based approach for the detection of swallow onsets is presented, which utilizes electromyography (EMG) and motion parameters of the larynx, measured by a wearable inertial measurement unit (IMU). For data collection, a pilot study with nine subjects was conducted, in which swallows with different volumes and consistencies, movement and speech were recorded. Evaluation of the detection approach is based on a cross-validation in combination with a grid search for finding appropriate thresholds. The detection approach results in a F1 score of 0.825 ± 0.074, rated as sufficient for a first feedback mechanism. This work supports the usage of the previously named signals as non-invasive measures for onset detection.