Segmentation of brain vessels from magnetic resonance angiography images

In this study, two stages were used for brain vessel segmentation from magnetic resonance images. Initially, brain tissue was separated from structures such as skull, eyes and nose. In order to efficient results, a method that fully automatically analyses the images inherently was used instead of a basic reference image called blind image analysis. Accordingly, the values of the filters utilized on magnetic resonance angiograms in the time-of-flight format were selected in a fully automatic mode, independent of the user. As a second step, brain vessel images were segmented over the segmented brain tissue. During this procedure, voxel density, vascular network neighbourhoods, and vascular network structure characteristics were evaluated. The obtained brain tissue and brain vessel images were compared with the manually segmented images using the dice coefficient method. As a result of the evaluation made on 45 data sets by finding an average of 100 frames in it, a 90.11% and 93.33% dice coefficient was obtained respectively for the brain tissue and brain vessels.