Surface-based parcellation and vertex-wise analysis of ultra high-resolution ex vivo 7 tesla MRI in Alzheimer's disease and related dementias

TL;DR

This study tackles the lack of scalable tools for ultra-high-resolution ex vivo MRI analysis in Alzheimer's disease and related dementias by assembling an 82-donor dataset imaged at 0.3 mm$^3$ isotropic resolution on 7T and developing a fast, automated surface-based pipeline. The pipeline integrates nnU-Net volumetric segmentation, CRUISE topology correction, and a FreeSurfer-based DKT atlas parcellation in native space to enable vertex-wise morphometry in template space, linking cortical thickness to histopathology measures. The results show significant region- and vertex-wise associations between reduced cortical thickness and AD pathology (amyloid-$\beta$, p-tau, neuronal loss, Braak, CERAD), with strongest effects in the medial temporal lobe, validating the approach for large-scale structure–pathology studies at ultra-high resolution. Open-source resources (dataset container, Jupyter notebooks) are provided to promote adoption and pave the way for ex vivo biomarkers to inform in vivo research.

Abstract

Magnetic resonance imaging (MRI) is the standard modality to understand human brain structure and function in vivo (antemortem). Decades of research in human neuroimaging has led to the widespread development of methods and tools to provide automated volume-based segmentations and surface-based parcellations which help localize brain functions to specialized anatomical regions. Recently ex vivo (postmortem) imaging of the brain has opened-up avenues to study brain structure at sub-millimeter ultra high-resolution revealing details not possible to observe with in vivo MRI. Unfortunately, there has been limited methodological development in ex vivo MRI primarily due to lack of datasets and limited centers with such imaging resources. Therefore, in this work, we present one-of-its-kind dataset of 82 ex vivo T2w whole brain hemispheres MRI at 0.3 mm isotropic resolution spanning Alzheimer's disease and related dementias. We adapted and developed a fast and easy-to-use automated surface-based pipeline to parcellate, for the first time, ultra high-resolution ex vivo brain tissue at the native subject space resolution using the Desikan-Killiany-Tourville (DKT) brain atlas. This allows us to perform vertex-wise analysis in the template space and thereby link morphometry measures with pathology measurements derived from histology. We will open-source our dataset docker container, Jupyter notebooks for ready-to-use out-of-the-box set of tools and command line options to advance ex vivo MRI clinical brain imaging research on the project webpage.

Figures (9)

• Figure 1: Schematic of the developed pipeline based on deep learning volumetric segmentations and surface-based modeling for parcellations of ex vivo whole hemisphere 0.3 mm$^3$ 7T MRI. The sequential steps follow A-F as described in Section 2.
• Figure 2: Ex vivo MRI segmentations and parcellations. Axial, coronal and sagittal viewing planes of ex vivo MRI at 0.3 mm$^3$ resolution for three subjects (A, B and C) with corresponding DKT volumetric segmentations and surface-based parcellations on pial and inflated surfaces for the medial and lateral views in native subject space resolution. Our method is able to correctly delineate the brain even in regions where the MR signal contrast is low in the anterior and the posterior brain MRI due to artifacts in acquisition protocol. Legend: See Fig. \ref{['fig:3']}.
• Figure 3: Spearman’s correlation plots between mean ROI thickness (mm) and neuropathological ratings in native subject-space. We observe significant negative correlation with global ratings of amyloid-$\beta$, Braak staging, CERAD, and the semi-quantitative ratings of the medial temporal lobe (MTL) neuronal loss and tau pathology. All the analysis were covaried for age, sex and postmortem interval (PMI) for the entire cohort of 82 subjects. See legend for the regional brain labels.
• Figure 4: Template-space vertex-wise morphometry-pathology correlations. Vertex-wise group analysis was performed to fit a generalized linear model (GLM). Shown are the statistical map (t-statistics) of the correlation between cortical thickness (mm) and with global ratings of amyloid-$\beta$, Braak staging, CERAD, and semi-quantitative ratings of the medial temporal lobe (MTL) neuronal loss and tau pathology, with age, sex and postmortem interval (PMI) as covariates across all 82 subjects. The clusters outlined in black indicate regions significant correlations (p$<$0.05) were observed after FWER correction for multiple comparisons.
• Figure A.1: Ex vivo tissue blockface photograph of a donor with co-morbid diagnosis of Parkinson’s disease and Lewy body disease (deceased at the age of 79). Shown are the lateral and the medial views of the right hemisphere.