A harmonized medial temporal lobe subregion segmentation protocol: an essential element for dementia research

The hippocampus and the parahippocampal region of the medial temporal lobe (MTL) consist of anatomically and functionally distinct subregions that are selectively targeted by different pathological processes in neurodegenerative disorders, such as Alzheimer’s disease (AD) and semantic variant Primary Progressive Aphasia (svPPA). In AD, in vivo MRI studies investigating MTL subregions have often shown mixed findings that are inconsistent with post-mortem pathology studies. These inconsistencies likely result from the lack of validity and standardization in MTL segmentation approaches. In svPPA, MTL subregions are severely affected, and automated tools fail to segment the subregions accurately. Sufficient spatial resolution of imaging voxels is required to visualize the internal MTL structures and landmarks that are critical for valid segmentation of the subregions. While more than 15,000 high-resolution scans have been collected worldwide, the lack of standardization among MTL subregional segmentation protocols is still a significant barrier. Variability in the definition of the MTL subregional boundaries among existing segmentation protocols have contributed to the conflicting findings in the field. This proposal seeks to accelerate progress made by our group, the Hippocampal Subfields Group, and complete our standardized, valid, and reliable harmonized protocol for the segmentation of MTL subregions in high-resolution MRI scans. We will leverage our ongoing, successful efforts, and evaluate the proposed harmonized protocol in scans from multicentric datasets. In Aim 1, we will finalize our harmonized segmentation protocol that is a) based on a large histology dataset and developed in collaboration with neuroanatomists, b) assessed by the larger community to reach consensus, c) tested for reliability, and d) implemented in an existing automated software program. In Aim 2, we will validate the harmonized protocol in early-stage AD through associations with markers of specific pathologies: a) test for anatomical specificity of the associations between atrophy rates across MTL subregions to PET measures of MTL tau-pathology and markers of cardiovascular risk in patients with Mild Cognitive Impairment and cerebral β-amyloid. In Aim 3, we will further evaluate the validity and utility of the harmonized protocol in svPPA by performing manual segmentation of the MTL subregions using the HSG protocol and testing for anatomical specificity of volume atrophy. In Aim 4, we will facilitate wide adoption of our protocol by providing training and education materials, in-person workshops for manual and automated segmentation, and making outcome measures publicly available. Our valid, reliable, and reproducible protocol for MTL subregions segmentation will benefit groups worldwide that have and continue to collect thousands of high-resolution data to study a variety of neurological and psychiatric conditions. It also lays the foundation for future research on how distinct disease processes in the MTL subregions may lead to impairments in specific cognitive functions. Our harmonization process can be adapted for achieving valid, harmonized segmentation for other brain regions.