Stylohyoid and Posterior Digastric Potential Evaluation for a Swallowing Detection With Intramuscular EMG

Total laryngectomy consists of the removal of the larynx and requires a tracheostomy to allow breathing, where the trachea is sewn on the anterior throat. This permanently separates the airway from the esophagus and any attempt to set it back in place would require to emulate the protective mechanism of the larynx, to avoid any aspiration during swallowing. So, with the aim to allow for the feasibility of an implantable active artificial larynx, and based on our previous investigations (Mialland et al., n.d.; 2023; 2023; 2022) this paper focuses on the stylohyoid and the posterior digastric muscles to evaluate their potential for a real-time and implantable swallowing detection algorithm. Indeed, we have shown that they activate at the beginning of swallowing, predominantly for swallowing, and that they are easily accessible, with no further impairment required during any sensor implantation. So, we measured them with intramuscular electromyography (EMG), along with the surface EMG of the submental muscles, to provide a basis for comparison. The swallowing sound was also measured with an accelerometer, to measure the vibration it generates through the skin, and define a temporal limit for the detection. 4 swallowing tasks and 13 non-swallowing tasks were investigated, to assess the muscles in terms of detection performances and abilities to provide early detection. Sub-groups of participants with similar swallowing, based on the Euclidean distance, were also explored. Thus, we found that the stylohyoid and the posterior digastric outperform the submental muscles and previously reported results in the literature, with increased performances when combined. Besides, the tasks that are the most related to the oral preparatory stage of swallowing tend to provide similar activity than the swallowing tasks, and are responsible for the most drastic drop in performances. Finally, even though we found promising results in terms of earliness, a real-time swallowing detection would benefit both from a stratified approach and from the development of an algorithm that is inherently built around a temporal constraint.