|M.Sc Student||Bord Asaf|
|Subject||Automatic Segmentation of MRI Images Emphasizing Separation|
of White and Gray Matter
|Department||Department of Biomedical Engineering||Supervisor||Professor Moshe Gur|
|Full Thesis text|
Brain tissue segmentation is a common practice in the field of brain research, for both anatomical and functional needs. Anatomical studies attempt to localize and quantify brain structures, while functional studies aim at understanding how these structures operate, independently and jointly. For these purposes, an accurate segmentation of the cortex, the "operational" layer of the brain, is necessary. Several aspects of functional imaging are instrumental only if they can be registered to an accurate anatomical model.
In this work a robust fully automatic method has been developed for extracting the White-matter - Gray-matter boundary, by employing a 3D segmentation algorithm. The CAG algorithm (Connectivity Analysis along Gray-scale levels) iteratively searches for significant blobs by applying increasing thresholds, and fine-tunes the resulting segments by optimizing a gradient-based energy measure along their edges. White-matter is thus extracted, the outer surface of which delineates the Gray-matter boundary. Testing the CAG algorithm on actual data yielded performance comparable with manual expert segmentation. When measured against common algorithms which tackle the same problem, the proposed algorithm proved comparable to most best performing ones, and better than a commercial solution which is commonly used. In addition, several tools for skull removal and ventricles' segmentation were used in conjunction with the main algorithm, to produce the final anatomical model, over which the functional data was laid.