Fused Segmentation of Geometric Models for Myocardium and Coronary Artery via Medial Axis

TL;DR

This paper introduces a novel medial axis-based algorithm for fused segmentation of coronary arteries and myocardium from cardiac CT models, enabling accurate identification of ischemic regions and supporting clinical decision-making.

Contribution

The study presents a new algorithm and software for segmenting coronary arteries and myocardium using medial axis computation, improving accuracy and efficiency over conventional methods.

Findings

Algorithm accurately segments coronary arteries and myocardium.

Provides robust linkage between myocardium-at-risk and arteries.

Experimentally verified with theoretical assertions.

Abstract

Coronary arteries and their branches supply blood to myocardium. The obstruction of coronary arteries results in significant loss of myocardium, called acute myocardial infarction, and the number one cause of death globally. Hence, quantification of the regional amount of heart muscle subtended by obstructed coronary arteries is of critical value in clinical medicine. However, the conventional methods are inaccurate and frequently disagree with clinical practices. This study proposes a novel algorithm to segment regional myocardium-at-risk subtended by any potentially obstructed coronary artery. Assuming the geometric triangular mesh models of coronary artery and myocardium derived from an individual cardiac computed tomography image, the proposed algorithm performs (i) computation of the medial axis of the coronary artery and (ii) segmentation of the coronary artery and myocardium using the medial axis. The algorithm provides the fused segmentation of coronary artery and myocardium via the medial axis. The computed result provides a robust mathematical linkage between myocardium-at-risk and supplying coronary arteries so that ischemic myocardial regions can be accurately identified, and both the extent and severity of myocardial ischemia can be quantified effectively and efficiently. Furthermore, the correspondence between segmented coronary artery and myocardium can be more importantly used for building optimization models of cardiac systems for various applications. We believe that the proposed algorithm and implemented VoroHeart program, which is freely available at http://voronoi.hanyang.ac.kr/software/voroheart, will be an invaluable tool for patient-specific risk predictions and the treatment of obstructed coronary artery disease in clinical medicine. The algorithmic accuracy and efficiency are theoretically asserted and experimentally verified.