Food Bolus Segmentation in Ultrasound Imaging for Dysphagia Rehabilitation

In recent years, there has been an increase in the incidence of aspiration pneumonia or choking caused by deteriorating swallowing function in an aging population. Evaluating swallowing function is crucial for the early detection and prevention of such conditions. However, standard methods for evaluating swallowing function are primarily invasive, posing challenges for regular rehabilitation. In this paper, we propose a dataset in ultrasound images focused on swallowing for visually interpretable food bolus detection. We also evaluate food bolus segmentation utilizing a U-net incorporating an Attention Gate to suppress feature activation in regions irrelevant for prediction. Experimental results on the swallowing dataset demonstrate the potential of our method for assessing swallowing in ultrasonography.