Uncertainty-Driven Edge Prompt Generation Network for Medical Image Segmentation
IEEE Transactions on Medical Imaging · 2025-10 · 44 (10) · 3950-3961
DOI: 10.1109/tmi.2025.3535478 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Transactions on Medical Imaging, 2025-10. Verfasst von Junyong Zhao und 6 weiteren Autor:innen. Hauptthemen laut MEDLINE: Image Processing, Computer-Assisted, Neural Networks, Computer. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 48 Literaturangaben. Gefördert von: National Natural Science Foundation of China, National Key Research and Development Program of China.
Abstract (Originalsprache)
Segment Anything Model (SAM) is a foundational image segmentation model, which shows superior performance for natural image segmentation tasks. Several SAM-based medical image segmentations have been proposed. However, these SAM-based medical image segmentation methods heavily depend on prior manual guidance involving points, boxes, and coarse-grained masks, which lack adaptability and flexibility. Moreover, the inherent challenge of edge blurring in medical images is critical, as it directly affects the quality of segmentation. To address these challenges, we propose an uncertainty-driven edge prompt generation network for medical image segmentation, called UDEG-Net. Specifically, to better adapt to medical image segmentation, we fine-tune the encoder by using Low-Rank Adaptation (LoRA) technology to enhance the encoder's learning capability and capture enriched medical image features. Furthermore, to overcome the limitations of interactive prompts, we develop an auto edge prompt generator to generate edge prompt information and further enhance the structural representation. Finally, to focus on the high-uncertainty edge areas, we introduce an evidence-based uncertainty estimation and a progressive uncertainty-driven loss to drive the auto edge prompt generator to yield robust edge prompt information and reliable segmentation results. Experimental results on three public datasets and one private dataset show that our UDEG-Net outperforms the state-of-the-art medical image segmentation methods.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Förderung
- National Natural Science Foundation of China
- National Key Research and Development Program of China
Zitieren
APA
Zhao, J., Sun, L., Fan, D., Wang, K., Si, H., Fu, H., & Zhang, D. (2025). Uncertainty-Driven Edge Prompt Generation Network for Medical Image Segmentation. IEEE Transactions on Medical Imaging, 44(10), 3950-3961. https://doi.org/10.1109/tmi.2025.3535478
Vancouver
Zhao J, Sun L, Fan D, Wang K, Si H, Fu H, et al. Uncertainty-Driven Edge Prompt Generation Network for Medical Image Segmentation. IEEE Transactions on Medical Imaging. 2025;44(10):3950-3961. doi:10.1109/tmi.2025.3535478
BibTeX
@article{Zhao2025,
author = {Zhao, Junyong and Sun, Liang and Fan, Dingwei and Wang, Kun and Si, Haipeng and Fu, Huazhu and Zhang, Daoqiang},
title = {Uncertainty-Driven Edge Prompt Generation Network for Medical Image Segmentation},
journal = {IEEE Transactions on Medical Imaging},
year = {2025},
volume = {44},
number = {10},
pages = {3950--3961},
doi = {10.1109/tmi.2025.3535478}
}
EUVIMED