Eskildsen, Simon Fristed3; Manjón, José V.2; Coupé, Pierrick2; Fonov, Vladimir2; Collins, D. Louis2
1 Department of Clinical Medicine - Center of Functionally Integrative Neuroscience, Department of Clinical Medicine, Health, Aarhus University2 unknown3 Department of Clinical Medicine - Center of Functionally Integrative Neuroscience, Department of Clinical Medicine, Health, Aarhus University
Brain cortical surface extraction from MRI has applications for measurement of gray matter (GM) atrophy, functional mapping, source localization and preoperative neurosurgical planning. Accurate cortex segmentation requires high resolution morphological images and several methods for extracting the cerebral cortex have been developed during the last decade (Dale 1999, Kim 2005, Eskildsen 2006). In many studies, the resolution of the morphological image acquisition sequence is chosen to be relatively low (~1mm3) due to time and equipment constraints. To improve segmentation accuracy, such low resolution images can be upsampled using various interpolation techniques. However, many interpolation methods lead to blurred images where high frequency information (e.g., edges) is badly reconstructed. To overcome this issue, superresolution methods have been proposed (Manjón 2010a), which have the ability to effectively increase the image resolution while preserving sharp features of the underlying anatomy. In this study, we investigated the effect of applying superresolution as proposed in (Manjon 2010a) to the accuracy of cerebral cortex segmentation.