Automated Tracking and Computer Reproduction of Vessels in DSA Images

Abstract

We are developing an automated vessel-tracking method based on the double-square-box region-of-search technique, for efficient tracking of the connected vascular tree in a digital subtraction angiography (DSA) image. Tracking points and branch vessels are located by searching of the perimeter of boxes, which are centered on previously determined tracking points. The most accurate results (90% true-positive rate with six false-positives per image) are obtained by tracking using the double-square-box method. In relatively straight regions of vessels, a large box is employed for efficient tracking; in curved regions of vessels, a small box is employed to ensure accurate tracking. When tracking is completed, accurate vessel information, ie, the vessel position, size, and contrast determined at each tracking point, is available for further quantitative analysis. Computer reproductions of tracked vessel trees appear to correspond well to those in DSA images.