Segmentation of thalamic nuclei based on tensorial morphological gradient of diffusion tensor fields

Although thalamic nuclei are not directly visible on conventional anatomical magnetic resonance images (MRI), it is possible to observe differences between the nuclei using diffusion tensor imaging (DTI), because of their distinct fiber orientation. This work presents a method to segment the various nuclei of human thalamus using diffusion MRI. Our approach is to use the watershed transform and other concepts from mathematical morphology to segment the nuclei. However, to segment structures using the tensor data produced with DTI (as opposed to scalar images) the concept of a tensorial morphological gradient (TMG) needs to be introduced. Based on the TMG, segmentation of the nuclei of the thalamus was successful using the watershed transform. Our segmentation is consistent with a histological atlas. Since the proposed method, as opposed to the majority of the DTI-based segmentation methods, does not require manual seed and/or surface placement, its results are highly repeatable.