A Wearable Fabric Sensing Device for Swallow Monitoring and Classification

Laryngeal elevation during swallowing is critical to airway protection, pharyngeal shortening, and upper esophageal sphincter opening. Currently, videofluoroscopy is the only way to quantify amplitude and timing of laryngeal elevation during swallowing. Although wearable devices exist, none are able to quantify amplitude of laryngeal elevation. This proof-of-concept study introduces a novel wearable fabric sensing device equipped with three knitted strain sensors, designed for real-time monitoring and precise classification of swallowing actions. Simultaneous recordings were made with our sensors and submental surface electromyography (sEMG) on 12 healthy adults who performed swallowing and non-swallowing tasks. K-nearest neighbors models were utilized to classify swallowing and non-swallowing tasks using data from knitted sensors and sEMG signals. Surveys were conducted to assess the comfort of wearing the device. The overall accuracy of KNN classifications to predict swallowing versus non-swallowing tasks was 0.97 for the knitted sensor, 0.82 for sEMG, and 0.98 for the combined dataset (knitted sensor + sEMG), while the accuracy to predict specific tasks was 0.75 for the knitted sensor, 0.32 for sEMG, and 0.77 for the combined dataset. Participants rated the knitted sensor's discomfort an average rating of 7.33/100, indicating a low level of discomfort. Our findings show that laryngeal movement during swallowing can be detected using knitted strain sensors worn on the outside of the neck, with the ability to distinguish between swallowing and non-swallowing tasks, as well as between types of swallows. The detection accuracy is significantly higher than that using the state-of-the-art sEMG method.