Precise Lumen Segmentation in Coronary Computed Tomography Angiography

Coronary computed tomography angiography (CCTA) allows for non-invasive identification and grading of stenoses by evaluating the degree of narrowing of the blood-filled vessel lumen. Recently, methods have been proposed that simulate coronary blood flow using computational fluid dynamics (CFD) to compute the fractional flow reserve non-invasively. Both grading and CFD rely on a precise segmentation of the vessel lumen from CCTA. We propose a novel, model-guided segmentation approach based on a Markov random field formulation with convex priors which assures the preservation of the tubular structure of the coronary lumen. Allowing for various robust smoothness terms, the approach yields very accurate lumen segmentations even in the presence of calcified and non-calcified plaques. Evaluations on the public Rotterdam segmentation challenge demonstrate the robustness and accuracy of our method: on standardized tests with multi-vendor CCTA from 30 symptomatic patients, we achieve superior accuracies as compared to both state-of-the-art methods and medical experts.