Vision-based Swallowing Capability Quantification with Smartphones

The diagnosis of dysphagia by clinicians poses challenges, as it requires patients to make regular clinic visits and the diagnosis requires training and empirical skills of the doctors. To address these issues and enable patients to assess potential symptoms of dysphagia easily at home, we propose a method to evaluate oral and swallowing function from videos captured by smartphones. To achieve this, we determined the required set of tasks for analyzing oral and swallowing function and analyzed videos of these tasks obtained from 147 participants. As a result, we obtained a discrimination performance with a balanced accuracy of 70.6% and a Weighted F1 score of 0.801.