Multi-graph Graph matching for coronary artery semantic labeling in invasive coronary angiograms

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Graph Theory Applications

Background:

• Coronary artery disease (CAD) is a leading global cause of mortality.
• Invasive coronary angiography (ICA) is the standard for evaluating coronary anatomy.
• Deep learning methods struggle with semantic labeling of coronary artery segments due to anatomical variations and branch similarities.

Purpose of the Study:

• To develop an advanced algorithm for accurate semantic labeling of coronary artery segments.
• To improve the detection of stenosis in coronary arteries.
• To leverage graph modeling for enhanced coronary artery analysis.

Main Methods:

• Modeled the coronary vascular tree as a graph.
• Proposed a multi-graph matching (MGM) algorithm incorporating cycle consistency.
• Utilized anatomical graph structure, radiomics features, and semantic mapping.

Main Results:

• Achieved 0.9471 accuracy for coronary artery semantic labeling on a multi-site dataset (718 ICAs).
• Reached 0.9155 overall accuracy for stenosis detection using semantically labeled arteries.
• Demonstrated the effectiveness of MGM in handling varying arterial anatomy and projection angles.

Conclusions:

• The MGM algorithm offers a novel and accurate approach to coronary artery semantic labeling.
• This method enhances the analysis of coronary arteries derived from multiple ICAs.
• The findings provide valuable insights for understanding vascular health and pathology in CAD.