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

Abstract

Magnetic resonance imaging (MRI) is the standard tool to image the human brainin vivo. 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 [3], BigBrain [4], Julich [5]), are currently the state of the art and provide exquisite 3D cytoarchitectural maps, but lack probabilistic labels throughout the whole brain. Here we presentNextBrain, 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, 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 an atlas withprobabilisticlabels throughout the whole brain. Further, we created a companion Bayesian tool for automated segmentation of the 333 ROIs in anyin vivoorex vivobrain MRI scan using theNextBrainatlas. We showcase two applications of the atlas: automated segmentation of ultra-high-resolutionex vivoMRI and volumetric analysis of brain ageing based on ∼4,000 publicly availablein vivoMRI scans. We publicly release the raw and aligned data (including an online visualisation tool), probabilistic atlas, and segmentation tool. By enabling researchers worldwide to analyse brain MRI scans at a superior level of granularity without manual effort or highly specific neuroanatomical knowledge,NextBrainwill accelerate our quest to understand the human brain in health and disease.

Molecular signatures of cortical expansion in the human fetal brain

“…47,170,171 In such studies, comparisons with in vivo neuroanatomy can only be fully realised through three-dimensional localisation of tissue samples within a common coordinate space. 62,68,70 To date, a limitation of this approach has been either the sampling of a narrow age range outside of key developmental periods 35,69 or, in developmental datasets, a lack of 3D spatial information 3 and relatively coarse anatomical sampling. 34 To fill this gap, we provide a new resource, μBrain, built upon existing open-source data, to allow researchers to map developmental neuroanatomy of the human fetal brain onto early histogenic processes using contemporaneous post mortem data.…”

“…In addition to the whole brain volume and cortical atlas, we created partial 3D reconstructions of ISH staining for 41 genes (see Methods; Figure 1g). Based on an average 41 tissue sections per gene (Table S3), semi-quantitative maps of gene expression revealed the tissue-and region-specific distributions of several genes, including caudal enrichment of the transcription factor EOMES in the subventricular zone, 15 and markers of neuronal migration (DCX 66 ) and synaptic transmission (GRIK2 67 ), in the cortical plate (Figure 1g; Figure S5) Existing histological brain atlases, including those of the adult human, 62,68,69 mouse, 70,71 and macaque 72 brains facilitate integration with other data modalities, including neuroimaging, and are amenable to advanced computational image analysis methods to extract quantitative measures of neuroanatomy across multiple scales. 73,74 Building upon existing resources, 3,4 we have created the μBrain atlas (Figure 1; Figure S4), a new and freely-available 3D volumetric model of the 21 PCW fetal brain at 150𝜇m resolution, accompanied by a set of n=20 cerebral tissue labels (Figure S4a-b); surface models of the cortical plate surface and cortical plate/subplate interface with n=29 cortical area labels (Figure 4c) and n=41 partial reconstructions of ISH expression data (Figure 5).…”

“…Existing histological brain atlases, including those of the adult human, 62,68,69 mouse, 70,71 and macaque 72 brains facilitate integration with other data modalities, including neuroimaging, and are amenable to advanced computational image analysis methods to extract quantitative measures of neuroanatomy across multiple scales. 73,74 Building upon existing resources, 3,4 we have created the μBrain atlas ( Figure 1; Figure S4 ), a new and freely-available 3D volumetric model of the 21 PCW fetal brain at 150 µ m resolution, accompanied by a set of n=20 cerebral tissue labels ( Figure S4a-b ); surface models of the cortical plate surface and cortical plate/subplate interface with n=29 cortical area labels ( Figure 4c ) and n=41 partial reconstructions of ISH expression data ( Figure 5 ).…”