Characterization of hippocampal subfields using histology-based annotated postmortem MRI: Lessons for in vivo segmentation II.

High-resolution in vivo magnetic resonance imaging (MRI) of hippocampal subfields is a rapidly advancing field due to their implication in cognition, disorder, and disease. Hippocampal subfield segmentation on in vivo MRI is generally guided by postmortem reference material, which has been limited by small sample sizes that preclude comprehensive characterization of subfield border locations and their variability. Addressing this, we characterized hippocampal subfield border variability in two ultra-high-resolution postmortem MRI datasets with combined annotated histological sections, including cases with and without dementia. We examined: 1) the order of appearance and disappearance of subfields along the long axis of the hippocampus; 2) the order of appearance and disappearance of subicular subregions; 3) the medial-lateral position of subicular subregional boundaries along the hippocampal body; 4) the location of the CA3 relative to hippocampal head digitations; 5) the subfield borders in the hippocampal body relative to a volume proportion of the dark band; and 6) the association of hippocampal length and subiculum-CA1 border location with diagnosis, demographic factors, and factors related to postmortem imaging. Our findings reaffirmed that there is a consistent order of appearance and disappearance of subfields in the hippocampal head and tail, respectively. The subicular subregions exhibited a 'first in, last out' order of appearance and disappearance, and pre/parasubiculum consistently occupied half of the subicular complex in coronal slices throughout the hippocampal body. Hippocampal head digitations were not a reliable landmark for CA3 appearance, but SRLM proportionality did offer a potentially consistent approach for estimating CA2 and CA3 subfield borders in relation to the hippocampal border. No clear relationship was found between the anatomical features and diagnosis, demographic factors, and factors related to postmortem imaging. These findings have implications for the development and harmonization of hippocampal subfield segmentation protocols and interpretation of high-resolution functional MRI studies of the human hippocampus.