Subcortical structure segmentation using probabilistic atlas priors

The segmentation of the subcortical structures of the brain is required for many forms of quantitative neuroanatomic analysis. The volumetric and shape parameters of structures such as caudate are employed to characterize a disease or its evolution. This paper presents our fully automatic segmentation of the caudate. The segmentation is based on an unbiased diffeomorphic atlas with probabilistic spatial priors built from a training set of MR images with corresponding manual segmentations. When applying the atlas for automatic structural segmentation, an MR image is first intensity inhomogeneity corrected, skull stripped and intensity calibrated to the atlas. Then the atlas image is registered to the image using an affine followed by a deformable registration matching the gray level intensity. Finally, the registration transformation is applied to the caudate probabilistic maps, which are then thresholded at 0.5 probability. Our method has been tested on all datasets provided by workshop as our atlas was build on a separate training population. The results show intermediate overlap results (76% Dice) and high correlation with the IBSR data (93%) and moderate correlation with the BWH data (64%). This indicates that our manual segmentation procedure is more similar to the procedure used for the IBSR than for the BWH dataset.