A next-generation, histological atlas of the human brain and its application to automated brain MRI segmentation

Adrià Casamitjana,Matteo Mancini,Eleanor Robinson,Loïc Peter,Roberto Annunziata,Juri Althonayan,Shauna Crampsie,Emily Blackburn,Benjamin Billot,Alessia Atzeni,Oula Puonti,Yaël Balbastre,Peter Schmidt,James Hughes,Jean C Augustinack,Brian L Edlow,Lilla Zöllei,David L Thomas,Dorit Kliemann,Martina Bocchetta,Catherine Strand,Janice L Holton,Zane Jaunmuktane,Juan Eugenio Iglesias

DOI: https://doi.org/10.1101/2024.02.05.579016

2024-09-06

Abstract:Magnetic resonance imaging (MRI) is the standard tool to image the human brain . In this domain, digital brain atlases are essential for subject-specific segmentation of anatomical regions of interest (ROIs) and spatial comparison of neuroanatomy from different subjects in a common coordinate frame. High-resolution, digital atlases derived from histology (e.g., Allen atlas [ ], BigBrain [ ], Julich [ ]), are currently the state of the art and provide exquisite 3D cytoarchitectural maps, but lack probabilistic labels throughout the whole brain. Here we present , a next - generation probabilistic atlas of human brain anatomy built from serial 3D histology and corresponding highly granular delineations of five whole brain hemispheres. We developed AI techniques to align and reconstruct ∼10,000 histological sections into coherent 3D volumes with joint geometric constraints (no overlap or gaps between sections), as well as to semi-automatically trace the boundaries of 333 distinct anatomical ROIs on all these sections. Comprehensive delineation on multiple cases enabled us to build . Further, we created a companion Bayesian tool for automated segmentation of the 333 ROIs in any or brain MRI scan using the atlas. We showcase two applications of the atlas: automated segmentation of ultra-high-resolution MRI and volumetric analysis of Alzheimer’s disease and healthy brain ageing based on ∼4,000 publicly available MRI scans. We publicly release: the raw and aligned data (including an online visualisation tool); the probabilistic atlas; the segmentation tool; and ground truth delineations for a 100 μm isotropic hemisphere (that we use for quantitative evaluation of our segmentation method in this paper). By enabling researchers worldwide to analyse brain MRI scans at a superior level of granularity without manual effort or highly specific neuroanatomical knowledge, holds promise to increase the specificity of MRI findings and ultimately accelerate our quest to understand the human brain in health and disease.

Neuroscience

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is to construct a high - precision probabilistic atlas of human brain anatomy to achieve automatic segmentation of human brain MRI scans. Specifically, the researchers developed NextBrain - a new - generation probabilistic atlas of human brain anatomy, which is based on multi - modal 3D histological data of five complete hemispheres and has been annotated with detailed regions. The construction of this atlas aims to overcome the limitations of existing atlases in terms of coverage, label density, and registration between MRI and histological images, thereby providing higher - resolution and more detailed brain region segmentation capabilities.

### Main Problems and Solutions:
1. **Improving the Level of Detail in MRI Analysis**:
- **Problem**: The resolution of existing MRI atlases is insufficient to study sub - structures in the brain (such as sub - regions of the hippocampus, thalamic nuclei, etc.), and these sub - structures have significant differences in function and connectivity.
- **Solution**: By constructing a probabilistic atlas based on high - resolution histological data, NextBrain can provide more fine - grained brain region segmentation, thereby improving the accuracy of MRI analysis.

2. **Achieving a Probabilistic Atlas across the Entire Brain**:
- **Problem**: Current histological atlases usually have limited coverage and lack probabilistic labels across the entire brain.
- **Solution**: NextBrain constructs a probabilistic atlas across the entire brain by densely annotating multi - modal 3D histological data of five complete hemispheres, which can be used for MRI segmentation of the entire brain.

3. **Solving the Challenges of MRI - Histological Image Registration**:
- **Problem**: There are technical difficulties in the registration between MRI and histological images, especially when dealing with a large number of non - parallel slices and tissue deformation.
- **Solution**: The researchers developed a customized image registration framework, combining AI technology and Bayesian optimization methods, ensuring high - precision registration between different - modality images.

4. **Automating the Segmentation of MRI Scans**:
- **Problem**: Manual segmentation of MRI scans is time - consuming and requires highly specialized neuroanatomical knowledge.
- **Solution**: NextBrain provides an accompanying Bayesian tool that can automatically segment 333 specific brain regions, greatly improving the segmentation efficiency and accuracy.

### Application Examples:
- **Automatic Segmentation of Ultra - High - Definition Ex Vivo MRI**: Demonstrates the application of NextBrain in processing ultra - high - resolution ex vivo MRI scans.
- **Volumetric Analysis of Alzheimer's Disease and Healthy Brain Aging**: Based on approximately 4,000 publicly available in vivo MRI scans, NextBrain can distinguish between the disease group and the control group and conduct detailed volumetric analysis.

Through these innovations, NextBrain not only improves the accuracy and efficiency of MRI analysis but also provides valuable resources for researchers, promoting the understanding of the human brain in healthy and diseased states.