Size independent active contour model for blood vessel lumen quantification in high-resolution magnetic resonance images

Abstract

Atherosclerosis is the most common cause of myocardial infarction. To study the atherosclerotic plaque in high resolution Magnetic Resonance images we developed a software tool called ATHER. An active contour model is used for segmentation and quantification of blood vessel lumen. Its implementation, based on a dynamic scaling process, presents two interesting features: 1) independence of the influence of the inflating force from the current size of the contour, 2) strong reduction of the computational cost. Therefore the contour converges very quickly even when initialized by a single point. This paper reports a validation of the model in ex vivo vascular images from Watanabe heritable hyperlipidaemic rabbits. Results of automatic quantification were compared to measurements performed by experts. Average difference of the area measurements between ATHER and the experts was equal to the inter­observer variability, but intra-variability of the automatic measurements was significantly smaller than the intra-observer variability.