DTI-based Corpus Callosum parcellation using the Tensorial Morphological Gradient and Self-Organizing Maps

The Corpus Callosum (CC) is a major white matter structure that plays a crucial role in the communication between the cerebral hemispheres and is associated with many subjects’ characteristics and brain diseases. The subdivision of the CC into smaller regions, called parcellation, is important for a detailed analysis of the structure. However, the lack of visible landmarks makes the CC parcellation a challenging task. In this study, we introduce a novel automatic data-driven approach based on Diffusion Tensor Imaging. It uses the Tensorial Morphological Gradient (TMG) and Self-Organizing Maps for CC parcellation into five regions. Qualitative and quantitative comparisons were performed between the proposed method and three other methods in the literature, using a dataset composed of images from five hundred subjects. Results demonstrate that our method provides distinct CC regions and exhibits similarities with one of the compared methods, which is also data-driven. Moreover, the proposed methodology is the only of the compared methods to consider the complete tensorial information, provided by the TMG map.