Answers are grounded in this paper and cite their sources.
Investigates food bolus segmentation in dysphagia rehabilitation, suggesting improved patient safety.
Dysphagia impairs the ability to eat and increases the risk of aspiration pneumonia and choking. Dysphagia rehabilitation often requires strengthening the muscles that transport residual food boluses from the pharynx to the stomach. Detecting food boluses in the pharynx is a crucial early indicator of aspiration risk during rehabilitation and is vital for patient safety. However, prior detection methods are invasive and rely on extensive equipment, making them unsuitable for routine clinical application. To address these limitations, we focus on food bolus segmentation in ultrasonography, which minimizes patient burden. In this paper, we propose a swallowing ultrasound image dataset designed to support supervised learning approaches for food bolus segmentation. It contains ultrasound images capturing the swallowing reflex with seven types of food, accompanied by pixel‐precise ground‐truth annotations for all food boluses. We also propose a novel segmentation framework to address image noise and class imbalance in the data. Specifically, we introduce attention gate modules to suppress irrelevant feature activation in noisy regions like the esophageal wall. Moreover, we employ a learning strategy targeting food bolus regions to mitigate class imbalance due to their small size. Extensive experiments demonstrate the effectiveness of our framework for food bolus segmentation in ultrasound images.
No takes yet. Share an insight, caveat, or question.
Gao et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: