Automated Phase-Based Segmentation of the Cerebral Cortex in 7T MR Images of the Elderly

Nhat Trung Doan¹, Maarten J. Versluis², Sanneke van Rooden, Jeroen van der Grond, Andrew Webb², Mark A. van Buchem, Johan H.C. Reiber¹, Julien Milles¹

The aim of this work is to implement a phase-based approach for the automated segmentation of the cerebral cortex in T2* data of elderly patients acquired using a 7T MR scanner. These data show essentially no gray/white matter contrast on magnitude images. The method is divided into two stages. The first step makes use of k-means clustering to segment the outer layer of the cortex. The obtained contour is subsequently deformed to match the gray/white matter interface based on the improved contrast in phase images, thereby resulting in a full segmentation of the cerebral cortex.