Automatic detection of coronary stenosis in X-ray angiography through spatio-temporal tracking

Abstract

Automatic detection of coronary stenosis in X-ray angiog-raphy data is a challenging problem. The low contrast between vessels and surrounding tissue, as well as large intensity gradients within the image, make detection of vessels and stenoses difficult. In this paper we exploit the spatiotemporal nature of the angiography sequences to present a robust method for automatically isolating the coronary artery tree. An arterial width surface is formed for each isolated artery segment by calculating the width along a segment and tracking the segment in each image frame over time. A persistent minima of this surface then corresponds to a stenosis in the artery. Results of testing on a variety of stenosis locations in various coronary arteries are presented and compared to stenosis detected from single frame analysis. This method is able to detect the presence of stenosis in an artery segment with a sensitivity of 86% and a specificity of 97% on 16 patients with a total of 20 image runs. This is the first fully automatic method for stenosis detection in X-ray angiography.