Towards automated coronary artery segmentation: A systematic review

Area of Science:

• Medical imaging analysis
• Computational anatomy
• Cardiovascular imaging

Background:

• Vessel segmentation is a critical initial step in processing 3D medical images for clinical and research applications.
• Current segmentation techniques are often manual, time-consuming, and impractical for large-scale datasets.

Purpose of the Study:

• To systematically review and analyze coronary artery segmentation methods published since 2010.
• To focus on fully automated methods suitable for the increasing volume of medical imaging data.
• To explore trends and characteristics of different segmentation approaches over the past decade.

Main Methods:

• Systematic literature review of manuscripts published from 2010 onwards.
• Categorization of segmentation methods based on their underlying approach.
• Analysis of trends and characteristics of various segmentation techniques.

Main Results:

• Segmentation methods were broadly categorized into region growing, voxelwise prediction, and partitioning approaches.
• Region growing, especially active contour models, was historically common but has declined.
• Convolutional neural networks have seen a significant rise in popularity.

Conclusions:

• Coronary artery segmentation is trending towards machine learning methods.
• There is a growing emphasis on developing efficient segmentation techniques.
• Computationally intensive methods like vesselness and multiplanar reformation are becoming less favored.