A computational framework for ultra-high resolution cortical segmentation at 7Tesla.

This paper presents a computational framework for whole brain segmentation of 7Tesla magnetic resonance images able to handle ultra-high resolution data. The approach combines multi-object topology-preserving deformable models with shape and intensity atlases to encode prior anatomical knowledge in a computationally efficient algorithm. Experimental validation on simulated and real brain images shows accuracy and robustness of the method and demonstrates the benefits of an increased processing resolution.