Model-Driven Centerline Extraction for Severely Occluded Major Coronary Arteries

Abstract

Almost all previous approaches on coronary artery centerline extraction are data-driven, which try to trace a centerline from an automatically detected or manually specified coronary ostium. No or little high level prior information is used; therefore, the centerline tracing procedure may terminate early at a severe occlusion or an anatomically inconsistent centerline course may be generated. In this work, we propose a model-driven approach to extracting the three major coronary arteries. The relative position of the major coronary arteries with respect to the heart chambers is stable, therefore the automatically segmented chambers can be used to predict the initial position of these coronary centerlines. The initial centerline is further refined using a machine learning based vesselness measurement. The proposed approach can handle variations in the length and topology of an artery, and it is more robust under severe occlusions than a data-driven approach. The extracted centerlines are already labeled, therefore no additional vessel labeling procedure is needed. Quantitative comparison on 54 cardiac CT datasets demonstrates the robustness of the proposed method over a state-of-the-art data-driven approach.