Diffusion MRI provides a wealth of information regarding the micro-structural integrity and connectivity of the human brain in vivo. High angular resolution diffusion imaging (HARDI) allows more accurate characterization of complex neuronal micro-architecture than conventional diffusion tensor imaging (DTI), and its once low scanning efficiency has recently become comparable to that of DTI. Therefore, HARDI is rapidly becoming a powerful research and clinical tool. Spatial normalization of HARDI information is necessary in a) whole-brain voxel-wise investigations, b) region-of-interest (ROI) studies comparing the underlying white matter fiber directions, and c) tract-specific HARDI investigations with automated atlas-based tract segmentation. The accuracy of such studies depends on the accuracy of spatial normalization. Accurate spatial normalization requires a high-quality HARDI template representative of the human brain. Study-specific templates can be constructed, but have several limitations. Furthermore, in the case of tract-specific HARDI investigations, accurate tract segmentation also requires that the HARDI template is accompanied by a set of semantic labels accurately delineating different white matter tracts (forming an accurate white matter atlas). Currently available white matter digital atlases are problematic, since they have been either generated based on anatomical landmarks, thus mixing tracts with substantially different roles, or using DTI tractography, which fails in regions with crossing fibers. Therefore, a detailed probabilistic brain atlas is needed, containing a) a high-quality HARDI template representative of the human brain, and b) accurate white matter labels generated with HARDI tractography. Since spatial misalignment can be detrimental in voxel-wise studies of white matter micro-structure, a number of investigations have adopted an alignment-invariant approach where information is first projected onto a white matter """"""""skeleton"""""""", followed by voxel-wise analysis on the skeleton. Tract-specific studies using atlas-based segmentation can also be negatively affected by atlas-to-subject misregistration, and could benefit from the alignment-invariant approach if a skeletonized version of an accurate white matter atlas was available. Finally, information on both white and gray matter is often combined in neuroimaging research, and although it is possible to spatially match independently generated white and gray matter atlases, anatomical correspondence between atlases is not guaranteed due to registration errors, as well as due to anatomical differences between the populations used for atlas construction. Therefore, the objective of this project is to develop an accurate, comprehensive white and gray matter probabilistic atlas of the adult human brain (both conventional and skeletonized), also containing multi-channel, artifact-free anatomical MRI and HARDI information. Successful completion of this work will bring forth a powerful set of tools that will increase the accuracy of both voxel-wise and tract-specific investigations of the adult human brain micro-structure. The results of this project will dramatically enhance the role of diffusion MRI as a diagnostic tool for a wide range of clinical problems.
Diffusion MRI provides a wealth of information regarding the micro-structural integrity and connectivity of the human brain in vivo. The objective of this project is to develop an accurate, comprehensive white and gray matter digital atlas of the adult human brain, also containing high-quality anatomical and diffusion MRI information. Successful completion of this work will bring forth a powerful set of tools that will increase the accuracy of both voxel-wise and tract-specific investigations of the adult human brain micro-structure.
