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

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.

• The paper presents a novel pipeline combining deep learning and classical methods for automated segmentation and surface-based parcellation.
• The approach achieves accurate 0.3 mm^3 resolution parcellation and reveals significant correlations between cortical thickness and neuropathology.
• It offers a scalable framework by publicly releasing both the high-resolution dataset and computational tools for future ADRD research.

Automated Surface-based Parcellation and Vertex-wise Analysis of High-resolution Ex Vivo 7 Tesla MRI in Neurodegenerative Diseases

Introduction

Neurodegenerative diseases, particularly Alzheimer's disease and related dementias (ADRD), pose significant diagnostic challenges. The heterogeneity of these conditions necessitates the use of high-resolution imaging to accurately detect co-pathologies. Ex Vivo Magnetic Resonance Imaging (MRI) at ultra high-resolution offers an unparalleled view of brain structure, potentially revealing details hidden in in vivo scans. Despite its potential, the methodological development for ex vivo MRI has been hampered by the scarcity of high-resolution datasets and the complexity of analyzing these data. This paper introduces a novel dataset of 82 ex vivo whole-brain hemispheres at 0.3 mm^3 resolution, spanning various neurodegenerative conditions. Leveraging this dataset, the authors developed a computational pipeline for automated surface-based parcellation of ex vivo brain tissue, facilitating structure-pathology association studies at an unprecedented resolution.

Dataset Description

The dataset comprises 82 ex vivo T2w whole-brain hemispheres MRI scans at 0.3 mm^3 resolution. These scans embody a diverse spectrum of ADRD diagnoses and were collected with adherence to ethical standards. The brain hemispheres underwent thorough histological examination to provide a comprehensive account of neuropathological markers, aiding in the accurate correlation with neuroimaging data.

Methodological Pipeline

The pipeline introduced is a novel amalgamation of deep learning and classical computational frameworks designed specifically for ultra-high resolution ex vivo MRI. The pipeline encompasses:

• Volume-based Segmentation: Utilizing a trained deep learning model to segment various brain regions in hemispheres, followed by a morphology-aware correction step to address complexities unique to ex vivo MRI.
• Surface-based Parcellation: Adapting FreeSurfer's surface-based methods to work with the ultra-high resolution data, enabling the delineation of cortical regions based on the Desikan-Killiany-Tourville (DKT) atlas.

This approach not only leverages deep learning for initial segmentation but also incorporates robust methods for topology correction and surface-based modeling, ensuring accurate anatomical parcellations in native subject space.

Experimental Results

• Segmentation and Parcellation: Key results include the successful application of the computational pipeline to the dataset, yielding accurate anatomical parcellations at 0.3 mm^3 resolution.
• Region-based Cortical Thickness vs. Neuropathology Correlations: Significant negative correlations were identified between cortical thickness in specific brain regions and various neuropathological markers, aligning with expectations based on known disease progression profiles.
• Surface-based Vertex-wise Cortical Thickness Correlations: The pipeline facilitated template-space analyses revealing significant associations between cortical thickness and neuropathology across the cortex.

Discussion

The study underscores the potential of combining advanced deep learning segmentation techniques with classical computational anatomy methods to analyze high-resolution ex vivo MRI. The approach demonstrated not only overcomes obstacles inherent to ex vivo MRI analysis but also sets a precedent for future large-scale studies aiming to elucidate structure-pathology relationships in neurodegenerative diseases.

Conclusion

The development of an automated surface-based parcellation pipeline for high-resolution ex vivo MRI represents a significant advancement in neuroimaging research. By applying this pipeline to a large-scale dataset of ADRD diagnoses, this work provides a foundation for in-depth structure-pathology association studies. Future developments may include the exploration of other anatomical atlases and the incorporation of additional MRI modalities to further refine our understanding of neurodegenerative diseases.

This research is a substantial step forward in the quest to utilize high-resolution ex vivo MRI for the detailed study of neurodegenerative diseases. By making their dataset, tools, and methodology publicly available, the authors significantly contribute to advancing the field of MRI-based neurological disorder research.