A semi-supervised fracture-attention model for segmenting tubular objects with improved topological connectivity
Yanfeng Zhou1,2, Liqun Zhong1,2, Zichen Wang1,2
1School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China.
Bioinformatics (Oxford, England)
|January 12, 2025
Summary
This study introduces a semi-supervised fracture-attention model (SSFA) to improve tubular object segmentation by enhancing connectivity and reducing fractures. SSFA offers a novel fracture rate metric for better evaluation, outperforming existing methods.
Area of Science:
- Medical image analysis
- Computer vision
Background:
- Accurate segmentation of tubular structures is vital for medical imaging analysis.
- Current deep learning methods struggle with connectivity and fractures, limiting scalability due to reliance on labeled data.
Purpose of the Study:
- To develop a semi-supervised model that enhances connectivity and reduces fractures in tubular object segmentation.
- To introduce a new metric for quantitatively assessing segmentation fractures.
Main Methods:
- Proposed a semi-supervised fracture-attention model (SSFA) for tubular object segmentation.
- Developed a novel evaluation metric, the fracture rate, to assess segmentation quality.
Main Results:
- SSFA demonstrated improved connectivity and reduced fractures compared to state-of-the-art methods.
- The model achieved superior topological performance across four public datasets.
- The fracture rate metric provided an intuitive quantitative assessment of segmentation fractures.
Conclusions:
- SSFA effectively addresses limitations in current tubular object segmentation, particularly regarding connectivity and fractures.
- The proposed fracture rate metric offers a valuable tool for evaluating segmentation quality.
- The developed model and metric advance the field of medical image analysis for tubular structures.
