Towards automatic identification of cerebral arteries in 4D flow MRI

Abstract

4D flow MRI is a powerful imaging technique which provides an angiographic image with information about blood flow in a large volume, time resolved over the cardiac cycle, in a short imaging time. This study aims to develop an automatic method for identification of cerebral arteries. The proposed method is based on an atlas of twelve arteries, developed from 4D flow MRI of 25 subjects. The atlas was constructed by normalizing all images to MNI-space, manually identifying the arteries and creating an average over the volume. The identification is done by extracting a vascular skeleton from the image, transforming it to MNI-space, labeling it with the atlas and transforming it back to subject space. The method was evaluated on a pilot cohort of 8 subjects. The rate of correctly identified arteries was 97%.