A Fast Automatic Method for 3D Volume Segmentation of the Human Cerebrovascular

We present a new method for 3-D volume segmentation of the human cerebrovascular structures from Magnetic Resonance Angiograms (MRA) and Magnetic Resonance Ventriculargrams (MRV). A slice through the volume containing large vein or artery structures is chosen, which becomes the seed location for the segmentation process. A modified 3-D computer graphics based region-filling algorithm is used to sweep the vascular tree from seed locations in order to track and label the vascular structure. The labeled-segmented images are extracted and a 3-D model is created using VTK toolkit. The 3-D model can then be viewed on a stereo graphics capable workstation or in a Virtual Reality environment. We also present a new maximum intensity projection (MIP) based technique for validating the results. We implemented MIP and used it to project both the segmented and original volume at different angles. Hence, the resultant images from both volumes can be compared side by side, which facilitates validation process.