A Visible Camera Approach to Motion Tracking Based Swallowing Evaluation

Video fluoroscopy (VF) is considered the gold standard in swallowing evaluation, however it involves exposure of X-rays and cannot be performed repeatedly. Therefore, we propose a method for analyzing the laryngeal area using a visible light camera as a non-invasive evaluation of swallowing. In this paper, we verify the effectiveness of this camera-based evaluation. First, we track the laryngeal area in GoPro and VF videos taken simultaneously using DeepLabCut. We create graphs from the extracted coordinate data, synchronize them, and then combine them into a single graph to confirm the degree of overlap of the waveform. Additionally, we use MATLAB to identify the peak, the start and end of swallowing, and the positions of the maximum and minimum gradients. We define four types of time intervals, such as the time from the start to the end of swallowing and the time from the maximum gradient to the end of swallowing and confirm the correlation between these intervals. As a result, we found that the waveforms on the graphs were generally consistent. We also found a correlation between the swallowing times. The coefficient of determination was 0.5 or higher, indicating a high accuracy of the analysis. Therefore, there is a correspondence between movements in the laryngeal area with GoPro and VF videos, and we showed that the evaluation for swallowing using a visible light camera is an effective method.