An Effective Brain Vasculature Segmentation Algorithm for Time-of-Flight MRA Data

Blood vessel segmentation is important for diagnosis of different vascular pathologies and planning of surgical treatments. In this study, we present an efficient approach for the brain vasculature extraction. The proposed method is based on multiview projection and an integrated active contour model. First, the magnetic resonance angiography (MRA) dataset is enhanced by Frangi filter and then projected from different angles to the 2D plane to make up the small area vessels occupy in each slice and to avoid overlapping and covering between vessels. Second, blood vessels on the maximum intensity projection (MIP) images are segmented by the integrated active contour model to extract as much low contrast and thin vessels as possible. Lastly, the corresponding voxels are selected by projecting back the segmented pixels, and then are used to construct the blood vessels in three-dimensional space. Experiments on 2D images and 3D volumes demonstrated that the proposed approach has superior efficiency and accuracy.