Improved segmentation of intracranial volumes using a combination of T1- and T2-weighted 3D MRI
Helms G, Kallenberg K, Dechent P
Gttingen University, Faculty of Medicine
Major problems of brain extraction and segmentation when using the FMRIB software library (FSL) can be overcome by T2-contrast features. T1w (MDEFT) and T2w (SPACE) structural 3D-MRI (1 mm resolution) were acquired at 3 Tesla. Postprocessing comprised rigid-body coregistration, brain extraction, segmentation, and volumetry. T1-contrast provided reliable segmentation of gray and white matter; T2-contrast delineation of the intracranial space and identification of a hypointense class of non-brain, representing vessels and connective tissue. Mismatches in sub-volumes determined from one single contrast could be largely explained by the 4-class results. This approach may improve assessment of cortical atrophy in ageing and dementia.