Related Experiment Video
Updated: Sep 10, 2025
04:48
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
Published on: July 5, 2024
491
DB-SNet: A dual branch network for aortic component segmentation and lesion localization.
Mingliang Yang1, Jinhao Lyu2, Jianxing Hu2
1School of Medical Technology, Beijing Institute of Technology, No.5 Zhongguancun South Street, Haidian District, Beijing 100081, China.
Summary
DB-SNet, a novel dual-branch network, accurately segments aortic components and lesions from CT angiography (CTA) scans. This efficient model significantly speeds up analysis, offering a valuable tool for cardiovascular diagnosis in clinical settings.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Cardiovascular Disease
Background:
- Accurate segmentation of aortic components (lumen, calcification, false lumen) and lesions (aneurysm, stenosis, dissection) in CT angiography (CTA) is critical for cardiovascular diagnosis and treatment planning.
- Existing automated methods often produce binary masks with limited clinical utility and require separate pipelines for anatomical and lesion segmentation, increasing resource demands.
Purpose of the Study:
- To introduce DB-SNet, a dual-branch 3D segmentation network based on MedNeXt, designed for efficient and integrated analysis of aortic structures and lesions in CTA scans.
- To improve upon the limitations of existing methods by enabling simultaneous segmentation of multiple aortic features with enhanced accuracy and reduced computational cost.
Main Methods:
- Developed DB-SNet, a dual-branch 3D segmentation network featuring a shared encoder and task-specific decoders, integrated with a novel channel-space cross-fusion module for improved feature interaction.
- Conducted a systematic ablation study to optimize backbone architectures, information interaction strategies, and loss weight configurations for dual-task performance.
- Evaluated the model on a large dataset of 435 multi-center CTA cases for training and 493 external cases for validation.
Main Results:
- DB-SNet outperformed 15 state-of-the-art models, achieving superior Dice Similarity Coefficient (DSC: 0.615) and Intersection over Union (IoU: 0.524) scores.
- Significantly reduced model parameters by 64.8% and computational complexity by 36.4% compared to the leading MedNeXt model.
- Achieved a remarkable 30.801x inference speedup (37.985s vs. 1170s for manual annotation), demonstrating high efficiency.
Conclusions:
- DB-SNet presents a new paradigm for efficient and integrated aortic analysis, balancing model efficiency with high accuracy.
- The proposed network offers a robust solution for real-time cardiovascular diagnosis, particularly in resource-constrained clinical environments.
- The developed dataset and code are publicly available to facilitate further research and clinical application.
Keywords:
Aortic component segmentationAortic lesion segmentationComputational efficiencyDual-branch segmentation network (DB-SNet)Thoracoabdominal CTA
