Coronary lumen and plaque segmentation from CTA using higher-order shape prior.

Area of Science:

• Medical image analysis
• Computer vision
• Computational anatomy

Background:

• Accurate segmentation of coronary lumen and plaque in CT angiography is crucial for cardiovascular disease assessment.
• Existing methods may struggle with complex vessel geometries and plaque heterogeneity.

Purpose of the Study:

• To develop a novel segmentation method for coronary lumen and plaque in CT angiography.
• To incorporate shape priors using higher-order potentials within a multi-label graph cut framework.

Main Methods:

• Utilized multi-label graph cuts with higher-order potentials to enforce shape priors.
• Defined higher-order potentials based on candidate shapes (tubular for vessel walls, globular for calcified plaques).
• Employed Hessian analysis for candidate shape detection and integrated corresponding higher-order terms into the energy function.

Main Results:

• The proposed method demonstrated effectiveness in segmenting coronary lumen and plaque.
• Achieved performance comparable to top methods in sensitivity and positive predictive value on the MICCAI 2012 challenge dataset.
• Secured a top average rank in the standardized evaluation framework.

Conclusions:

• The novel segmentation approach effectively leverages shape priors through higher-order potentials.
• This method offers a robust solution for coronary lumen and plaque segmentation in CT angiography.
• The technique shows significant potential for clinical applications in cardiovascular imaging.