Related Experiment Video
Updated: Jun 30, 2025
08:32
Adapting Human Videofluoroscopic Swallow Study Methods to Detect and Characterize Dysphagia in Murine Disease Models
Published on: March 1, 2015
21.3K
PECI-Net: Bolus segmentation from video fluoroscopic swallowing study images using preprocessing ensemble and
Dougho Park1, Younghun Kim2, Harim Kang2
1Pohang Stroke and Spine Hospital, Pohang, Republic of Korea; School of Convergence Science and Technology, Pohang University of Science and Technology, Pohang, Republic of Korea.
Computers in Biology and Medicine
|March 15, 2024
Summary
PECI-Net improves bolus segmentation in videofluoroscopic swallowing studies (VFSS) using novel preprocessing and cascaded inference techniques. This network enhances image quality and accuracy for detecting swallowing disorders.
Area of Science:
- Medical Imaging
- Computer Vision
- Biomedical Engineering
Background:
- Accurate bolus segmentation is vital for diagnosing swallowing disorders via videofluoroscopic swallowing studies (VFSS).
- VFSS images present challenges like low contrast, unclear boundaries, and translucency, hindering precise segmentation.
- Existing models struggle with the inherent limitations of VFSS image data.
Purpose of the Study:
- To develop an advanced network architecture, PECI-Net, for improved bolus segmentation in VFSS images.
- To address the difficulties posed by low-quality VFSS images in automated swallowing disorder detection.
Main Methods:
- Proposed PECI-Net, integrating a preprocessing ensemble network (PEN) and a cascaded inference network (CIN).
- PEN adaptively combines preprocessing algorithms to enhance VFSS image sharpness and contrast.
- CIN utilizes contextual information through asymmetric cascaded inference to refine bolus segmentation and mitigate errors.
Main Results:
- PECI-Net significantly outperformed four recent baseline models in bolus segmentation accuracy.
- Achieved a 4.54% improvement over TernausNet and a 10.83% improvement over UNet.
- Ablation studies validated the effectiveness of both PEN and CIN components in enhancing performance.
Conclusions:
- PECI-Net offers a robust solution for accurate bolus segmentation in VFSS analysis.
- The proposed PEN and CIN techniques effectively overcome VFSS image quality limitations.
- PECI-Net demonstrates potential for advancing automated detection of swallowing disorders.
