Abstracts

Rationale: Hippocampal sclerosis (HS) is a common cause of medically intractable epilepsy and recent data suggest several subtypes of HS exist. Recent advances with magnetic resonance imaging (MRI) have made visualization of hippocampal subfields possible and segmentation of these sub-structures potentially feasible.  However, recently released data suggest very poor inter-observer reliability with respect to histological definition of hippocampal subfields. These data raise fundamental questions regarding the histological "ground-truth" validity of current approaches to subfield segmentation. Recent studies have utilized curved multiplanar reconstruction (CMPR) to visualize hippocampal anatomy with MRI, but this method has not yet been applied to histological data.

Methods: We utilized the BigBrain data, an open-source database of serially sectioned histological data for our analyses. We hypothesized that CMPR would allow reliable histological delineation of subfields throughout the hippocampal long axis (IE throughout the head, body, and tail). We aimed to examine the histological validity of our previously published protocol (Steve, 2017) (https://doi.org/10.1016/j.neuroimage.2017.06.008) throughout the hippocampal head, body, and tail.  In order to test these hypotheses, CMPR histology images were segmented at an interval of nine months by a single author (DH) for evaluation of intra-rater reliability.  The same images were then segmented by a second author (AA) for evaluation of inter-rater reliability.  Finally, segmentation according to our previous protocol (Steve 2017) was compared to ground truth histological measurements to evaluate the histological validity of our measurements.

Results: While conventional rectilinear visualization of histology revealed a complex hippocampal internal architecture (Figure 1), CMPR allowed histological visualization in a classic C-shape throughout the hippocampal long axis.  Intra- and inter-rater reliability was assessed via DICE overlap with the results shown in Table 1.

Conclusions: Curved multiplanar reconstruction (CMPR) can be applied to hippocampal histology and demonstrates that the characteristic 'C-shape’, seen with conventional rectilinear slicing only in the hippocampal body, is in fact maintained throughout the entire long axis of the human hippocampus. CMPR allows reliable subfield segmentation based on histological criteria throughout the hippocampal head, body, and tail. Our previous protocol (Steve 2017), which was initially developed in an independent sample of six hippocampal bodies, can be applied to provide histologically valid subfield measurements throughout the entire hippocampal long axis. These measurements have potential direct relevance to the measurement of hippocampal subfields in patients with medically intractable epilepsy due to HS.

Funding: Canadian Institutes of Health Research (CIHR) - Funding Reference Number (FRN): 183923