SACH-Net: Shape-Adaptive Convolution and Hierarchical Topology Constraints Framework for Coronary Artery Segmentation

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Cardiovascular Disease Diagnosis

Background:

• Accurate coronary artery segmentation is vital for diagnosing and planning treatment for coronary artery disease (CAD).
• Existing methods struggle with the complex, tree-like structure of coronary arteries, leading to segmentation errors like discontinuity and mis-segmentation of small branches.
• These limitations hinder efficient CAD management and clinical decision-making.

Purpose of the Study:

• To develop a novel framework, SACH-Net, for enhanced automatic coronary artery segmentation.
• To address the challenges of discontinuity and mis-segmentation in current coronary artery segmentation techniques.
• To improve the accuracy and reliability of coronary artery masks for clinical applications.

Main Methods:

• Proposed SACH-Net framework incorporating shape-adaptive convolution (SA-Conv) to learn vessel features adaptively.
• Introduced hierarchical topology constraints (HTC) to ensure continuity, overlap, and topological correctness in segmentation.
• Evaluated performance on the public ARCADE dataset.

Main Results:

• SACH-Net significantly outperformed state-of-the-art methods in coronary artery segmentation accuracy.
• The shape-adaptive convolution effectively captured intricate vascular structures.
• Hierarchical topology constraints successfully reduced segmentation fractures and discontinuities.

Conclusions:

• SACH-Net offers a significant advancement in automatic coronary artery segmentation.
• The framework demonstrates improved accuracy and robustness, holding promising clinical implications for medical image analysis.
• This research contributes to more effective computer-aided diagnosis and treatment planning for CAD.