High-Density Electromyography of Swallowing and Phonation: Methods for Automated Analysis

Dysphagia and dysphonia arise from dysfunctional neuromuscular coordination and are often under-diagnosed due to reliance on biomechanical assessment rather than electrophysiology. A non-invasive, high-density electromyography array and automated signal-processing framework was developed to capture neuromuscular activation patterns in swallowing and phonation. The custom flexible array (128 electrodes), positioned in under 10 minutes, recorded EMG from hyoid muscles in six healthy participants. The framework integrated audio markers to define functional onset and synchronize EMG events, enhancing temporal alignment. Window lengths were optimized using mutual information, and four artifact-exclusion methods were compared. The Local Outlier Factor has the highest performance, removing poor-quality channels with an F1 of 0.97 and specificity of 0.96. Regional signals were spatially averaged to characterize temporal myoelectrical signatures and extract morphological metrics. Temporal activation patterns were observed, with regional coordination across tasks. Swallowing had an activity duration of 0.7 ± 0.1 s, while phonation onset occurred at 0.3 ± 0.1 s. Two-dimensional maps visualized activation distribution and were projected onto 3D participant-specific geometries. Swallowing showed localization to the suprahyoid region, while phonation spanned both regions. The array effectively captured high-resolution myoelectrical signals, and the framework accurately visualized temporal and spatial signatures of dynamic coordination in swallowing and phonation.