Abstract
Structural changes of subregions in the medial temporal lobe (MTL) reflect the progression of Alzheimer’s disease. T2-weighted (T2w) MRI provides detailed depiction of the hippocampus and is the main \textit{in vivo} modality for MTL subregion segmentation. However, the existing \textit{in vivo} segmentation protocol relies on geometric rules rather than cytoarchitectonic criteria, which limits the anatomical accuracy of the resulting subregions. To address this, a cytoarchitecture-based protocol can be defined by mapping cytoarchitectonic annotations from histology to \textit{in vivo} MRI. The main challenge of this mapping lies in the anisotropic resolution of T2w MRI, on which the continuous histology-defined boundaries become fragmented. In this work, we employ an implicit neural representation to construct a continuous \textit{in vivo} representation of the MTL, onto which histology-defined boundaries are recovered smoothly. An isotropic segmentation method for fine-grained subregions is then developed with a coarse-to-fine two-stage framework. Compared with the anisotropic model trained in the original T2w space, the isotropic model achieves higher Dice scores in cross-validation. On the test set, the anisotropic model shows severe detection failures on fine-grained subregions, while the proposed model achieves the highest detection rate. Imaging biomarkers extracted from the proposed segmentation also achieve higher AUC values in early-affected AD subregions, in distinguishing amyloid-positive mild cognitive impairment from amyloid-negative cognitively unimpaired participants, compared to the other methods. This study shows that a continuous representation can effectively recover the anatomical structure of the MTL \textit{in vivo}, lowering the barrier to using cytoarchitectonic boundaries for \textit{in vivo} applications.
Code is \href{https://github.com/liyue3780/HyperResASHS/tree/twostage}{available}.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/Off_Grid_012.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: https://openreview.net/forum?id=ZP2mpcLPya
BibTex
@InProceedings{LiYue_From_MICCAISAT2026,
author = { Li, Yue AND Denning, Amanda E. AND Khodakarami, Zahra AND Khandelwal, Pulkit AND Xie, Long AND Brown, Christopher A. AND Wisse, Laura E. M. AND Das, Sandhitsu R. AND Wolk, David A. AND Yushkevich, Paul A.},
title = { { From Histology to in vivo MRI: An Implicit Neural Representation Framework for Medial Temporal Lobe Subregion Segmentation } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17277},
month = {pending},
page = {pending}
}