Framework for the analysis of the corpus callosum microstructure along its extension

Abstract

The corpus callosum is of great interest for the medical and research community, and its characteristics have been associated with many psychological disorders and brain diseases. Localized analysis of its features is a usual procedure, particularly for the diagnosis of multiple sclerosis and other inflammatory diseases. In this work, we propose a framework for extracting microstructure features along the corpus callosum extent into a signature function, allowing global and localized analyses to be performed in the 1-D domain of the signature, instead of the 3-D domain of the original image. Our solution is a succession of several specialized methods, which were designed to solve specific parts of the signature generation pipeline, including defining a plane of local symmetry for the corpus callosum internal fibers, perform the corpus callosum segmentation, trace the structure median axis, and extract the features along the median axis. A dataset with images from 80 distinct acquisitions from healthy subjects was used to evaluate both, the fiber's symmetry plane, and the generated signatures. Results show that the plane predicted by our method is significantly distinct from the planes predicted by traditional mid-sagittal plane estimation methods, with a larger difference on the inclination relative to the axial plane, of about 2 degrees on average. The signatures present a similar pattern in most cases but retain individual characteristics. The signatures generated by our proposed framework provide an unprecedented way to perform the analysis of the corpus callosum microstructure features, which is inherently localized, and independent from the structure morphology. Our solution open new possibilities for future related research and development in the field.