A postmortem computational atlas of the human hippocampus

The primary contribution of this dissertation is a computational atlas of the human hippocampus from high-resolution, postmortem magnetic resonance imaging (MRI) and densely-acquired histology. The atlas describes how the locations of subregion boundaries, derived from microscopic features extracted from histology, vary relative to structures and boundaries that can be seen in MRI. This work represents a major step towards understanding the hippocampal region\u27s anatomical variability by capturing anatomy of the region from the macroscopic to the microscopic level. The atlas reference space is generated from 0.2 mm isotropic resolution 9.4 telsa MRI of 26 whole-hippocampal specimens. A new groupwise diffeomorphic registration framework is developed that combines shape- and intensity-based volumetric registration of the MRI. Shape correspondences are imposed using a novel three-dimensional (3D), smooth spherical parameterization of the hippocampus. The parameterization is based on segmentations of both the whole hippocampus boundary and of the internal layers of the stratum radiatum and stratum lacunosum-moleculare (SRLM), and it characterizes hippocampal shape in a geometrically and anatomically intuitive manner. The hippocampal subregions in the atlas are derived from manual segmentation of the histological sections of eight atlas specimens acquired with 0.2 mm spacing, 5 or 7μm slice thickness, and stained using the Kluver-Barrera method. The atlas MRI template serves as the anatomical reference space for histological reconstruction and for visualization of the subregion labels. The histology is reconstructed manually using an interactive software tool that enables intuitive navigation and multi-scale viewing of histology and volumetric MRI together in 3D. The software\u27s multi-scale viewing capability permits fine-scaled alignment of specific regions of interest between slides and MRI. The feasibility of reconstructing maps of cytoarchitectonic features from the histology of one subject into the atlas space is also demonstrated. This is the first time that such a large collection of histology of the whole hippocampus has been acquired, labeled, reconstructed, and mapped into a common analysis space