Related Experiment Video
Updated: May 29, 2025
Visualization of Vascular and Parenchymal Regeneration after 70% Partial Hepatectomy in Normal Mice
Published on: September 13, 2016
VSNet: Vessel Structure-aware Network for hepatic and portal vein segmentation
Jichen Xu1, Anqi Dong2, Yang Yang3
1National Key Laboratory of Human-Machine Hybrid Augmented Intelligence, National Engineering Research Center for Visual Information and Applications, and Institute of Artificial Intelligence and Robotics, Xi'an Jiaotong University, Xi'an, China.
A new AI model, Vessel Structure-aware Network (VSNet), accurately segments liver vessels in CT scans, improving preoperative planning by capturing fine vein details. This method enhances segmentation of hepatic and portal veins, outperforming existing models.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Computer Vision
Background:
- Accurate segmentation of hepatic and portal veins from CT scans is vital for liver surgery planning.
- Current segmentation models often fail to capture the intricate details of smaller hepatic and portal veins, impacting clinical utility.
Purpose of the Study:
- To introduce Vessel Structure-aware Network (VSNet), a novel multi-task learning model designed for precise segmentation of hepatic and portal veins.
- To address the limitations of existing models in capturing fine vascular details and maintaining topological correctness.
Main Methods:
- Development of VSNet, a multi-task learning model incorporating a vessel-growing decoder.
- Creation and release of the largest dataset to date for hepatic and portal vessel segmentation, comprising 303 cases.
- Comparative experiments to evaluate VSNet's performance against other leading segmentation models.
Main Results:
- VSNet achieved superior Dice scores: 0.824 for hepatic veins and 0.807 for portal veins on the proposed dataset.
- The model demonstrated significant improvement over existing popular segmentation models.
- VSNet successfully captured topological features of minor veins and preserved connectivity to major vessels.
Conclusions:
- VSNet offers a significant advancement in the automated segmentation of hepatic and portal veins from CT scans.
- The model's ability to preserve vessel connectivity and capture fine details is crucial for enhanced preoperative planning.
- The publicly available dataset and source code facilitate further research and development in medical image analysis.
