Mapping how neurons are structurally wired into whole-brain networks can be challenging, particularly in larger brains where 3D microscopy is not available. Multi-modal datasets combining MRI and microscopy provide a solution, where high resolution but 2D microscopy can be complemented by whole-brain but low-resolution MRI. However, there lacks unified approaches to integrate and jointly analyse these multi-modal data in an insightful way. To address this gap, we introduce a data-fusion method for hybrid MRI-microscopy fibre orientation and connectome reconstruction. Specifically, we complement precise “in-plane” orientations from microscopy with “through-plane” information from MRI to construct 3D hybrid fibre orientations at resolutions far exceeding that of MRI whilst preserving microscopy's myelin specificity, resulting in superior fibre tracking. Our method is openly available, can be deployed on standard 2D microscopy, including different microscopy contrasts, and is species agnostic, facilitating neuroanatomical investigation in both animal models and human brains.

Biophysical modelling of diffusion MRI (dMRI) is used to non-invasively estimate microstructural features of tissue, particularly in the brain. However, meaningful description of tissue requires many unknown parameters, resulting in a model that is often ill-posed. The Bayesian EstimatioN of CHange (BENCH) framework was specifically designed to circumvent parameter fitting for ill-conditioned models when one is simply interested in interpreting signal changes related to some variable of interest. To understand the biological underpinning of some observed change in MR signal between different conditions, BENCH predicts which model parameter, or combination of parameters, best explains the observed change, without having to invert the model. BENCH has been previously used to identify which biophysical parameters could explain group-wise dMRI signal differences (e.g., patients vs. controls); here, we adapt BENCH to interpret dMRI signal changes related to continuous variables. We investigate how parameters from the dMRI standard model of white matter, with an additional sphere compartment to represent glial cell bodies, relate to tissue microstructure quantified from histology. We validate BENCH using synthetic dMRI data from numerical simulations. We then apply it to ex-vivo macaque brain data with dMRI and microscopy metrics of glial density, axonal density, and axonal dispersion in the same brain. We found that (i) increases in myelin density are primarily associated with an increased intra-axonal volume fraction and (ii) changes in the orientation dispersion derived from myelin microscopy are linked to variations in the orientation dispersion index. Finally, we found that the dMRI signal is sensitive to changes in glial cell load in the brain white matter, though no single parameter in the extended standard model was able to explain this observed signal change.

The BigMac dataset combines extensive post-mortem MRI with histology and polarised light imaging (PLI) in a macaque brain. Adding a new extension to our recently developed platform, the Tensor Image Registration Library (TIRL), we address unique challenges associated with this dataset and successfully align 77 2-D polarised light micrographs to high-resolution structural MRI data.

Acquisition of MRI and histology in the same ex-vivo tissue sample enables direct correlation between MR and histologically-derived metrics. Here, we analysed immunohistochemistry images of human visual cortex, anterior cingulate and hippocampus to produce stained area fraction maps for myelin, neurofilaments and microglia. We performed voxelwise correlations between MR parameters (FA, MD, R2*, R1) and histology maps to generally characterise the strength of relationships. We then used partial correlation to identify the unique variance in MR parameters explained by each histological feature, and multiple linear regression to explore how well multiple microstructural properties can together explain MR parameters.

Background Accurate registration between microscopy and MRI data is necessary for validating imaging biomarkers against neuropathology, and to disentangle complex signal dependencies in microstructural MRI. Existing registration methods often rely on serial histological sampling or significant manual input, providing limited scope to work with a large number of stand-alone histology sections. Here we present a customisable pipeline to automate the registration of stand-alone histology sections to whole-brain MRI data. Methods Our pipeline registers stained histology sections to whole-brain post-mortem MRI in 4 stages, with the help of two photographic intermediaries: a block face image (to undistort histology sections) and coronal brain slice photographs (to insert them into MRI space). Each registration stage is implemented as a configurable stand-alone Python script using our novel platform, Tensor Image Registration Library (TIRL), which provides flexibility for wider adaptation. We report our experience of registering 87 PLP-stained histology sections from 14 subjects and perform various experiments to assess the accuracy and robustness of each stage of the pipeline. Results All 87 histology sections were successfully registered to MRI. Histology-to-block registration (Stage 1) achieved 0.2-0.4 mm accuracy, better than commonly used existing methods. Block-to-slice matching (Stage 2) showed great robustness in automatically identifying and inserting small tissue blocks into whole brain slices with 0.2 mm accuracy. Simulations demonstrated sub-voxel level accuracy (0.13 mm) of the slice-to-volume registration (Stage 3) algorithm, which was observed in over 200 actual brain slice registrations, compensating 3D slice deformations up to 6.5 mm. Stage 4 combined the previous stages and generated refined pixelwise aligned multi-modal histology-MRI stacks. Conclusions Our open-source pipeline provides robust automation tools for registering stand-alone histology sections to MRI data with sub-voxel level precision, and the underlying framework makes it readily adaptable to a diverse range of microscopy-MRI studies. Highlights New software framework for prototyping bespoke image registration pipelines Automated pipeline to register stand-alone histology sections to whole-brain MRI Novel deformable slice-to-volume registration algorithm No strict necessity for serial histological sectioning for MRI-histology registration

Combining histology and ex vivo MRI from the same mouse brain is a powerful way to study brain microstructure. Mouse brains prepared for ex vivo MRI are often kept in storage solution for months, potentially becoming brittle and showing reduced antigenicity. Here, we describe a protocol for mouse brain dissection, tissue processing, paraffin embedding, sectioning, and staining. We then detail registration of histology to ex vivo MRI data from the same sample and extraction of quantitative histological measurements.

We introduce the Digital Brain Bank (open.win.ox.ac.uk/DigitalBrainBank), a digital platform providing open access to curated, multimodal post-mortem neuroimaging datasets. Datasets span three themes; Digital Anatomist: datasets for neuroanatomical investigations; Digital Brain Zoo: datasets for comparative neuroanatomy; Digital Pathologist: datasets for neuropathology investigations. The first release includes 21 distinctive whole-brain diffusion MRI datasets, alongside microscopy and complementary MRI modalities. This includes one of the highest-resolution whole-brain human diffusion MRI datasets ever acquired, whole-brain diffusion MRI in 14 non-human primate species, and one of the largest post-mortem whole-brain cohort imaging studies in neurodegeneration. Our resource facilitates incorporating post-mortem data into neuroimaging studies.

Understanding brain structure and function often requires combining data across different modalities and scales to link microscale cellular structures to macroscale features of whole brain organisation. Here we introduce the BigMac dataset, a resource combining in vivo MRI, extensive postmortem MRI and multi-contrast microscopy for multimodal characterisation of a single whole macaque brain. The data spans modalities (MRI and microscopy), tissue states (in vivo and postmortem), and four orders of spatial magnitude, from microscopy images with micrometre or sub-micrometre resolution, to MRI signals on the order of millimetres. Crucially, the MRI and microscopy images are carefully co-registered together to facilitate quantitative multimodal analyses. Here we detail the acquisition, curation, and first release of the data, that together make BigMac a unique, openly-disseminated resource available to researchers worldwide. Further, we demonstrate example analyses and opportunities afforded by the data, including improvement of connectivity estimates from ultra-high angular resolution diffusion MRI, neuroanatomical insight provided by polarised light imaging and myelin-stained histology, and the joint analysis of MRI and microscopy data for reconstruction of the microscopy-inspired connectome. All data and code are made openly available.

The acquisition of MRI and histology in the same post-mortem tissue sample enables direct correlation between MRI and histologically-derived parameters. However, there still lacks a standardised automated pipeline to process histology data, with most studies relying on manual intervention. Here, we introduce an automated pipeline to extract a quantitative histological measure for staining density (stain area fraction, SAF) from multiple immunohistochemical (IHC) stains. The pipeline is designed to directly address key IHC artefacts related to tissue staining and slide digitisation. Here, the pipeline was applied to post-mortem human brain data from multiple subjects, relating MRI parameters (FA, MD, RD, AD, R2*, R1) to IHC slides stained for myelin, neurofilaments, microglia and activated microglia. Utilising high-quality MRI-histology co-registrations, we then performed whole-slide voxelwise comparisons (simple correlations, partial correlations and multiple regression analyses) between multimodal MRI- and IHC-derived parameters. The pipeline was found to be reproducible, robust to artefacts and generalisable across multiple IHC stains. Our partial correlation results suggest that some simple MRI-SAF correlations should be interpreted with caution, due to the co-localisation of other tissue features (e.g., myelin and neurofilaments). Further, we find activated microglia-a generic biomarker of inflammation-to consistently be the strongest predictor of high DTI FA and low RD, which may suggest sensitivity of diffusion MRI to aspects of neuroinflammation related to microglial activation, even after accounting for other microstructural changes (demyelination, axonal loss and general microglia infiltration). Together, these results show the utility of this approach in carefully curating IHC data and performing multimodal analyses to better understand microstructural relationships with MRI.

Article Figures and data Abstract Editor's evaluation Introduction Results Discussion Materials and methods Appendix 1 Appendix 2 Appendix 3 Appendix 4 Data availability References Decision letter Author response Article and author information Metrics Abstract Post-mortem magnetic resonance imaging (MRI) provides the opportunity to acquire high-resolution datasets to investigate neuroanatomy and validate the origins of image contrast through microscopy comparisons. We introduce the Digital Brain Bank (open.win.ox.ac.uk/DigitalBrainBank), a data release platform providing open access to curated, multimodal post-mortem neuroimaging datasets. Datasets span three themes—Digital Neuroanatomist: datasets for detailed neuroanatomical investigations; Digital Brain Zoo: datasets for comparative neuroanatomy; and Digital Pathologist: datasets for neuropathology investigations. The first Digital Brain Bank data release includes 21 distinctive whole-brain diffusion MRI datasets for structural connectivity investigations, alongside microscopy and complementary MRI modalities. This includes one of the highest-resolution whole-brain human diffusion MRI datasets ever acquired, whole-brain diffusion MRI in fourteen nonhuman primate species, and one of the largest post-mortem whole-brain cohort imaging studies in neurodegeneration. The Digital Brain Bank is the culmination of our lab's investment into post-mortem MRI methodology and MRI-microscopy analysis techniques. This manuscript provides a detailed overview of our work with post-mortem imaging to date, including the development of diffusion MRI methods to image large post-mortem samples, including whole, human brains. Taken together, the Digital Brain Bank provides cross-scale, cross-species datasets facilitating the incorporation of post-mortem data into neuroimaging studies. Editor's evaluation This paper describes a new open-access digital brain bank of post-mortem brains that have been scanned with high-resolution, multimodal magnetic resonance imaging and with select datasets accompanied by histological data. This valuable resource can be used to study healthy human brains, pathological human brains, and the brains of other species, opening new opportunities for comparative neuroanatomy and the biological validation of non-invasive neuroimaging signals. https://doi.org/10.7554/eLife.73153.sa0 Decision letter Reviews on Sciety eLife's review process Introduction Magnetic resonance imaging (MRI) occupies a unique position in the neuroscience toolkit. In humans, MRI is used at the single-subject level diagnostically and is increasingly deployed at the population level in epidemiology (Marcus et al., 2007; Miller et al., 2016; Snoek et al., 2021; Van Essen et al., 2013). MRI is well-established in the context of imaging causal manipulations in experimental organisms ranging from mice (Denic et al., 2011; Thiessen et al., 2013) to nonhuman primates (Absinta et al., 2017; Klink et al., 2021) and provides precise measurements in cellular and tissue preparations (Wilhelm et al., 2012). This extensive landscape of overlap with the broader neuroscience toolkit creates the potential for MRI to facilitate integration between technologies and investigations. Although MRI hardware and acquisition protocols often need to be tailored to a specific domain, the underlying technology associated with all MRI measurements gives rise to a common set of signal forming mechanisms, facilitating cross-domain comparisons. There are few methods available to neuroscientists that span this breadth of domains. One challenge to the use of MRI as a bridging technology is the need for common measurements — for example, the same MRI measurements made across multiple species, or MRI and microscopy measurements in the same brain tissue (Mars et al., 2021). Post-mortem MRI provides unique opportunities for such common measurements. MRI in post-mortem tissue can be used to identify the origins of image contrast through integration with microscopy (Keren et al., 2015; Langkammer et al., 2012; Mollink et al., 2017), directly addressing concerns over the nonspecificity of MRI signals. In this context, post-mortem MRI data are important because they share common signal forming mechanisms with in vivo MRI and a common tissue state with microscopy, providing a framework for investigation across multiple spatial scales. Post-mortem MRI facilitates comparative anatomy investigations in species that are not traditionally accessible for in vivo imaging (Berns et al., 2015; Bhagwandin et al., 2017; Grewal et al., 2020; Heuer et al., 2019), including extinct species (Berns and Ashwell, 2017). Long post-mortem scans provide the opportunity to push the boundaries of spatial resolution, providing whole human brain coverage reaching voxel sizes of 100–500 μm (Edlow et al., 2019; Foxley et al., 2016; Fritz et al., 2019; Weigel et al., 2021), edging closer to microscopy techniques but benefitting from compatibility with in vivo imaging. As a nondestructive technique, post-mortem MRI enables the examination of tissue microstructure whilst preserving tissue, facilitating repeat MRI measurements with novel contrasts and technologies; and more generally, its integration with tools for post-mortem investigations (e.g., histopathology or proteomics). In this work, we introduce the Digital Brain Bank (open.win.ox.ac.uk/DigitalBrainBank), a data release platform resulting from a decade of post-mortem MRI research at the University of Oxford. The Digital Brain Bank provides open access to several post-mortem neuroimaging datasets spanning investigations into human neuroanatomy, cross-species neuroanatomy, and neuropathology. All datasets provide post-mortem MRI, including diffusion MRI, with complementary microscopy data (e.g., immunohistochemistry or PLI) included with some datasets. Our post-mortem imaging research has been specifically aimed at achieving whole-brain post-mortem MRI to support the investigation of multiple brain systems/regions and long-range connections (Foxley et al., 2014; Miller et al., 2011; Miller et al., 2012), and the first release to the Digital Brain Bank contains 21 distinct whole-brain post-mortem MRI datasets, including from whole human brains. All datasets are available to access, and prospective users of the Digital Brain Bank can explore a subset of data directly on the Digital Brain Bank website using Tview: a bespoke, open-source, and web-based image viewer. Tview has been developed for efficient browsing of imaging data spanning drastically different spatial scales, from submicron resolution microscopy to millimeter MRI acquisitions. It enables real-time visualization and interaction (zooming/panning) of both MRI and microscopy images, and with flexible overlays of different modalities. All datasets uploaded to the Digital Brain Bank are associated with researchers at the University of Oxford, or from close collaborators. Limited Derived Outputs from users of Digital Brain Bank datasets will also be considered for data upload. The first release to the Digital Brain Bank includes data from multiple published projects covering a breadth of neuroimaging research, including whole-brain diffusion MRI in 14 nonhuman primate species (Bryant et al., 2021; Roumazeilles et al., 2020; Roumazeilles et al., 2021), and one of the largest post-mortem whole-brain cohort imaging studies combining whole-brain MRI and microscopy in human neurodegeneration (Pallebage-Gamarallage et al., 2018). In addition, we present a previously unpublished project providing one of the highest-resolution whole-brain human diffusion MRI datasets ever acquired (500 μm isotropic resolution). The Digital Brain Bank will continue to grow over the coming years, with a number of further datasets already at the early stages of curation (Howard et al., 2019a; Martins-Bach et al., 2021; Martins-Bach et al., 2020; Wu et al., 2021). Results The Digital Brain Bank is accessible at open.win.ox.ac.uk/DigitalBrainBank. Datasets have been organized into categories reflecting three predominant themes of post-mortem neuroimaging research: Digital Anatomist: datasets for detailed neuroanatomical investigations. Digital Brain Zoo: datasets for comparative neuroanatomy. Digital Pathologist: datasets for neuropathology investigations. Here, we provide an overview of each theme, with examples from available datasets in the first release to the Digital Brain Bank. A brief description of all the datasets provided with the first release, alongside relevant publications, is provided in Table 1. Table 1 Description of all datasets provided in the first release to the Digital Brain Bank. All Structural MRI datasets in the first release were acquired using a balanced SSFP (bSSFP) or T2-weighted sequence, which yields strong gray-white matter contrast in formalin-fixed post-mortem tissue. Diffusion MRI datasets were acquired using a combination of diffusion-weighted steady-state free precession (DW-SSFP) and diffusion-weighted spin-echo (DW-SE) sequences. Full details of the motivation behind the choice of sequences and available contrasts are described in the Discussion. †T2* and magnetic susceptibility maps are currently available in 9 out of 12 ALS brains and all control brains. The remaining datasets were either lost during scanner export, or are of insufficient data quality for public release. CategoryNameContents: MRIResolution (MRI)Contents: MicroscopyRelevant publicationsDigital AnatomistHuman High-Resolution Diffusion MRI-PLIWhole-brain diffusion MRI, structural MRI, quantitative T1 and T2 maps:– Control human brain: 1×Diffusion MRI: (500 μm, 1 and 2 mm iso.)Structural MRI: 312.5×312.5×500 μm3T1 map: (0.75×0.75×1.6 mm3)T2 map: (0.75×0.75×1.6 mm3)Polarised light imaging (4 μm in-plane) in the anterior commissure, corpus callosum, pons, thalamus, and visual cortex (same brain)Dataset described in this publication (Methodology in Appendix 1), Diffusion MRI processing described in Tendler et al., 2020b, T2 mapping described in Tendler et al., 2021Digital AnatomistHuman Callosum MRI-PLI-HistologyCorpus callosum diffusion MRI:– Excised control human corpus callosum samples: 3×Diffusion MRI: (400 μm iso.)Polarised light imaging (4 μm in-plane), bright-field microscopy images of immunohistochemistry stains (0.25 μm in-plane) for PLP (myelin) and GFAP (astrocyte) (same human corpus callosum samples)Mollink et al., 2017Whole-brain diffusion MRI and structural MRI (available in brains marked with a *):Bushbaby (Galago senegalensis): 1×Capuchin monkey (Sepajus apella): 1×Chimpanzee* (Pan troglodytes): 2×Colobus monkey (Colobus guereza): 1×Cotton-Top tamarin (Saguinus oedipus): 1×Golden Lion tamarin (Leontopithecus rosalia): 1×Hamadryas baboon* (Papio hamadryas): 1×Macaque monkey (Macaca mulatta): 3×Mangabey (Lophocebus albigena): 1×Night monkey, (Aotus lemurinus): 1×Ring-tailed lemur (Lemur catta): 3×Saki monkey (Pithecia pithecia): 1×Western Lowland gorilla* (Gorilla gorilla): 1×Woolly monkey (Lagothrix lagotricha): 1×Diffusion MRI:300 μm iso.: Bushbaby, Cotton-Top tamarin & Golden Lion Tamarin400 μm iso: Night monkey500 μm iso: Ring-tailed lemur and Saki monkey600 μm iso: Capuchin monkey, Chimpanzee, Colobus monkey, Hamadryas baboon, Macaque monkey, Mangabey, Western Lowland Gorilla and Woolly MonkeyStructural MRI200 μm iso: Western Lowland Gorilla220 μm iso: Hamadryas Baboon0.22×0.22×0.19 mm3: 1× Chimpanzee0.375×0.375×0.40 mm3: 1× ChimpanzeeNone1× Western Lowland gorilla and 1× Chimpanzee described in Roumazeilles et al., 2020, 3× Macaque monkey and 3× Ring-Tailed Lemur described in Roumazeilles et al., 2021. Hamadryas baboon, Cotton-Top tamarin and Golden Lion tamarin datasets described in this publication (Methodology in Appendix 1). All other datasets described in Bryant et al., 2021Digital Brain ZooMarsupialsWhole-brain diffusion MRI and structural MRI:Tasmanian devil (Sarcophilus harrisii): 2×Thylacine (Thylacinus cynocephalus): 2×Diffusion MRI:1 mm iso: 1× Tasmanian devil1.5 mm iso: 1× Tasmanian devil1.1 mm iso: 1× Thylacine1.0×1.1×0.8 mm3: 1× ThylacineStructural MRI330 μm iso: 1× Tasmanian devil and 1× Thylacine330×330×300 μm3: 1× Tasmanian devil500 μm iso: 1× ThylacineNoneBerns and Ashwell, 2017Digital Brain ZooCetaceansWhole-brain diffusion MRI and structural MRICommon dolphin (Delphinus delphis): 1×Pantropical dolphin (Stenella attenuata): 1×Diffusion MRI: (1.3 mm iso.)Structural MRI: (640×640×500 μm3)NoneBerns et al., 2015Digital Brain ZooCarnivoraWhole-brain diffusion MRI and structural MRI:– European wolf (Canis lupus): 1×Diffusion MRI: (600 μm iso.)Structural MRI: (220 μm iso.)NoneDataset described in this publication (Methodology in Appendix 1)Digital PathologistHuman ALS MRI-HistologyWhole-brain diffusion MRI, structural MRI, quantitative T1, T2, and T2* maps, magnetic susceptibility maps (selected brains†):Amyotrophic lateral sclerosis (ALS) human brains: 12×Control human brains: 3×Diffusion MRI: (850 μm iso.)Structural MRI: (230–250 μm in-plane; 270–500 μm slice)T1 map: (0.65–1 mm in-plane; 0.90–1.6 mm slice)T2 map: (0.65–1 mm in-plane; 0.90–1.6 mm slice)T2*/magnetic susceptibility maps: (0.5 mm in-plane; 1.1–1.3 mm slice)Bright-field microscopy immunohistochemistry stains (0.50 μm in-plane, exception pTDP43 – 0.25 μm in-plane): pTDP-43, IBA1 (pan microglia), CD68 (activated microglia/macrophages), PLP (myelin), SMI-312 (axonal phosphorylated neurofilaments), and ferritin (iron storage, subset of regions)Regions: Anterior cingulate cortex, corpus callosum, hippocampus, primary motor cortex, and visual cortex (same brains).Selected multimodal histology available in two brains (1× ALS and 1× Control), and multiregional PLP (available in 10 out of 12 ALS brains and all control brains, 5–8 regions per brain) in first data release – remaining histology being actively curated.Pallebage-Gamarallage et al., 2018, Magnetic susceptibility and T2* mapping protocol described in Wang et al., 2020, Diffusion MRI processing described in Tendler et al., 2020b, T2 mapping described in Tendler et al., 2021 Digital Anatomist Datasets within the Digital Anatomist provide a new direction for answering fundamental questions in neuroanatomy, through ultra-high resolution MRI data and complementary microscopy within the same sample in humans and model nonhuman species. The long scan times available in post-mortem MRI affords imaging at ultra-high spatial resolutions, facilitating the delineation of small tissue structures within the human brain, one of the key aims of the Digital Anatomist. Often, post-mortem investigations are limited to small sections of excised brain tissue that represent a limited anatomical region. However, our developments in whole-brain post-mortem diffusion imaging (Foxley et al., 2014; McNab et al., 2009; Miller et al., 2011; Miller et al., 2012; Tendler et al., 2020b) provide the opportunity to investigate structural connectivity and gross neuroanatomy, at scales that are unobtainable in vivo. These developments have culminated in the Human High-Resolution Diffusion MRI-PLI dataset, providing one of the highest-resolution whole-brain human diffusion MRI datasets ever acquired (500 μm isotropic resolution), as shown in Figure 1. Companion datasets acquired at 1 mm and 2 mm (isotropic) provide a comparison at cutting-edge and conventional in vivo resolutions (Figure 1a). Figure 1 Download asset Open asset The Digital Anatomist. (a) Whole-brain diffusion MRI data available in the Human High-Resolution Diffusion MRI-PLI dataset reveals the wealth of information provided at increased spatial scales, one of the key aims of the Digital Anatomist. Here, the 500 μm dataset uncovers the information hidden at lower spatial resolutions, for example, visualizing the interdigitating transverse pontine fibers with the corticospinal tract or striations through the internal capsule. (b) Similarly, datasets across multiple spatial scales can inform us of the limitations at reduced imaging resolutions. Here, gyral tractography (occipital lobe) reveals an overall pattern of fibers turning into the gyral bank at 0.5 mm. At 1 mm, an underestimation of connectivity at the gyral banks is observed, known as the 'gyral bias' (Cottaar et al., 2021; Schilling et al., 2018). At 2 mm, tractography bears little resemblance to the expected architecture. Multimodal comparisons enable us to validate our findings, with complementary polarised light imaging (PLI) data at over 2 orders of magnitude increase in resolution (125×) revealing a similar pattern of gyral connectivity, and (c) excellent visual agreement with tractography across the pons. (a) displays diffusion tensor principal diffusion direction maps (modulated by fractional anisotropy). In addition to providing a new insight into human neuroanatomy, these data can be used to inform experimental design and the interpretation of results. Here, the Human High-Resolution Diffusion MRI-PLI dataset enables users to identify the resolution required to visualize certain brain structures (Figure 1a), and how spatial resolution impacts tractography performance (e.g., overcoming 'gyral bias'—Figure 1b; Cottaar et al., 2021; Schilling et al., 2018). Polarised light imaging (PLI) provides estimates of myelinated fiber orientation (Axer et al., 2011), and complementary PLI data acquired in a subset of brain regions (4 μm in-plane) facilitates cross-scale comparisons (Figure 1b and c). A further aim of the Digital Anatomist is to perform quantitative validations across modalities, relating MRI to microscopic measures. These kinds of analyses can only be achieved with accurately coregistered data, enabling pixel-wise comparisons across modalities acquired at drastically different spatial resolutions. This potential is most clearly seen in the Human Callosum MRI-PLI-Histology dataset, which provides diffusion MRI (400 μm isotropic), alongside complementary PLI (4 μm in-plane), and histology (myelin and astrocytes) (0.25 μm in-plane) in three excised human corpus callosum samples (Mollink et al., 2017). These data offer multiple pathways of investigation, including the identification of the origins of image contrast; validation of microstructural models of tissue (Mollink et al., 2017); and developing unique models explicitly linking MRI with microscopy (Howard et al., 2019b). Digital Brain Zoo The Digital Brain Zoo provides curated datasets to investigate neuroanatomy in nonhuman species and compare anatomy across species. Post-mortem MRI has enormous potential to inform comparative neuroanatomy for three reasons. First, it enables the scanning of species that would be extremely difficult or impossible to study in vivo. Second, samples can be imaged with minimal handling and without invasive procedures, enabling the study of rare specimens that would not be appropriate to dissect. Third, MRI investigations can be performed in whole-brain samples, rather than excised tissue sections. This makes post-mortem MRI ideally placed to characterize macroscopic brain structure, long-range structural connectivity, and tissue microstructure in species that are not traditional experimental models, and in particular rare species where very few brain samples may be available (Berns and Ashwell, 2017; Bhagwandin et al., 2017; Grewal et al., 2020; Mars et al., 2014). MRI data from multiple species allows one to formally compare brain organization, important for large-scale comparative neuroscience which has traditionally relied on very limited measures (e.g., whole or regional brain size measures of brain organization) (Mars et al., 2014). The ability to acquire data from whole brains opens up the possibility of elucidating principles of neural diversity across larger orders of mammalian species (Friedrich et al., 2021), and create between-species mappings to formally identify homologies and quantify unique aspects of any given brain (Mars et al., 2018). This also allows one to improve translational neuroscience by better understanding the relationship between the human brain and that of model species (e.g., macaque, marmoset, rat, and mouse). The Digital Brain Zoo provides access to post-mortem imaging datasets in nonhuman species covering multiple taxonomic ranks (Figure 2a), including nonhuman primate species (Bryant et al., 2021; Roumazeilles et al., 2020; Roumazeilles et al., 2021), Carnivora (Grewal et al., 2020), Marsupials (Berns and Ashwell, 2017), and Cetaceans (Berns et al., 2015). As with other collections in the Digital Brain Bank, the Digital Brain Zoo currently focuses primarily on whole-brain diffusion MRI. These datasets offer multiple pathways of investigation in comparative neuroanatomy, for example, through the examination of structural connections across brains (Figure 2b; Bryant et al., 2021). Furthermore, our developments in imaging large post-mortem samples have enabled us to acquire several high-quality post-mortem imaging datasets in species with brains that are too large to fit into specialized preclinical MRI systems, conventionally used to improve image quality in post-mortem MRI (see Discussion). Figure 2 Download asset Open asset The Digital Brain Zoo. (a) The first release of the Digital Brain Zoo provides whole-brain MRI datasets spanning multiple species and taxonomic ranks. Notably, we provide whole-brain diffusion MRI datasets from 14 nonhuman primate species, with samples selected for their high quality and to ensure sampling of all major branches of the primate evolutionary tree (Prosimian, New World monkey, Old World monkey, and Great Ape). (b) compares the relative volume of four tracts derived from nine nonhuman primate post-mortem datasets provided in the Digital Brain Zoo (Bryant et al., 2021), where increased distance from the centre corresponds to an increased volume. Digital Pathologist Datasets within the Digital Pathologist provide a new direction for examining neuropathology and MRI-pathology correlates in humans and established laboratory models. One of the biggest challenges in the use of MRI clinically is the lack of specificity to disease mechanisms. Many neurological diseases are characterized by changes at the cellular and subcellular level, which cannot be directly visualized with the limited resolution of MRI. Nevertheless, MRI contrast can be made sensitive to cellular-level phenomena that are relevant to disease. Acquisition of MRI and histology in the same tissue enables us to relate microscopic changes in the neural microenvironment to MRI image contrast. The primary aim of the Digital Pathologist is to facilitate these cross-scale comparisons, imaging brain tissue associated with a neurological disease. Such data are provided in the Human ALS MRI-Histology dataset (Figure 3a), which aims to identify how neuropathological changes in amyotrophic lateral sclerosis (ALS) give rise to altered MRI contrast, and answer specific questions related to ALS pathology. The Human ALS MRI-Histology dataset provides whole-brain multimodal MRI and selective histology in a cohort of 12 ALS (diagnosis during lifetime, confirmed ALS neuropathology) and 3 control (no known neuropathology) brains (Pallebage-Gamarallage et al., 2018) provided by the Oxford Brain Bank. MRI data includes diffusion, structural, quantitative susceptibility maps (via quantitative susceptibility mapping, QSM), and quantitative T1, T2, and T2* maps. Histology includes markers for proteinopathy (pTDP-43), microglia (CD68 and IBA1), myelin (PLP), neurofilaments (SMI-312), and iron (ferritin) in order to detect changes in a range of microstructures within cortical and subcortical regions (anterior cingulate cortex, corpus callosum, hippocampus, primary motor cortex, and visual cortex) associated with different proposed stages of ALS disease progression (Jucker and Walker, 2013). Figure 3 Download asset Open asset The Digital Pathologist. One of the key aims of the Digital Pathologist is the examination of neuropathological spread in neurological disease. The Human ALS MRI-Histology dataset (a) facilitates these investigations, combining whole-brain multimodal MRI and histology (selected brain regions) in a cohort of 12 ALS and 3 control brains. (b) Displays the reconstruction of five white matter pathways associated with different ALS stages in a single post-mortem brain (Kassubek et al., 2014). Comparisons between ALS and control brains over the corpus callosum of the cohort (c) reveals changes in fractional anisotropy (FA, normalized to Par/Temp/Occ lobe), with biggest changes associated with motor and prefrontal regions (Hofer and Frahm, 2006) (*=p<0.05; **=p<0.05 following multiple comparison correction) (full details of the corpus callosum analysis provided in Appendix 2). This reflects the anticipated changes in ALS with brain regions associated with motor function, in good agreement with a previous study (Chapman et al., 2014), which identified the greatest FA difference between ALS and controls in these regions. Accurate MRI-histology coregistrations facilitate cross-modality comparisons, and (d) displays an example of MRI-histology coregistration over the visual cortex of a single ALS brain achieved using the Tensor Image Registration Library (TIRL) (Huszar et al., 2019). V1=principal diffusion direction, FA=fractional anisotropy, MD=mean diffusivity, D⊥=radial diffusivity, MO=mode from diffusion tensor output, Dyad1=principal dyad orientation, f1=principal fiber fraction and D=diffusivity from Ball and Two-Stick output, swMRI=susceptibility-weighted MRI, χ=magnetic susceptibility. Details of stain contrasts in (b) and (d) are provided in Table 1. ALS, amyotrophic lateral sclerosis; MRI, magnetic resonance imaging. Different MRI modalities have known sensitivities to different components of the cellular environment. Combined with multimodal histology, these data provide the opportunity to relate neuropathologically induced changes in tissue microstructure to MRI image contrast. While these aims could be partially achieved by dissecting and scanning subregions of the brain, our approach of scanning whole post-mortem brains enables us to investigate neuropathological spread across the entire brain (Jucker and Walker, 2013). This facilitates investigations across long-range fiber-tracts associated with pathology (Figure 3b), or microstructural changes in multiple brain regions (Figure 3c). Notably, these analyses are being facilitated by accurate cross-modality image coregistrations (Huszar et al., 2019), enabling us to perform pixel-wise evaluations and integrate structural analyses to identify how pathology influences MR image contrast (Figure 3d) in a subset of brain regions associated with different proposed stages of ALS disease progression (Jucker and Walker, 2013). MRI data for the Human ALS MRI-Histology dataset for all 12 ALS and 3 control brains are immediately available to access, alongside a subset of histology data. Remaining histology data and MRI-histology coregistrations are being actively curated for future release to the Digital Brain Bank. Tview The Digital Brain Bank website enables users to browse a subset of data easily. A key feature of many datasets is that they contain both MRI and microscopy data. Few available viewers, whether downloadable or online, support both MRI and microscopy file formats, creating a barrier of entry for potential users. Moreover, 2D microscopy datasets are extremely high resolution: single images can exceed 10,000,000,000 pixels, running to gigabytes in size. We aim to provide online viewing of microscopy and MRI data using standard internet browsers, with a viewer that can handle data at very different spatial scales, provide flexible image overlays, and support color visualization for diffusion-derived measures, PLI, and multiple histological counterstains. Unable to identify existing software with these features, we developed a web-based image viewer, Tview. Tview is based on software originally used to display satellite imagery at multiple elevations, and enables real-time visualization, interaction (zooming/panning), and flexible overlays of different modalities in a single 2D plane of MRI and microscopy data. Visualization of multimodal (i.e., MRI and microscopy) datasets on the Digital Brain Bank website is achieved with Tview. An example Tview implementation is available at open.win.ox.ac.uk/DigitalBrainBank/#/tileviewer, where cross-modality coregistrations were performed using the Tensor Image Registration Library (TIRL) (Huszar et al., 2019) and FNIRT (Andersson et al., 2007; Jenkinson et al., 2012), both available as part of FSL. Code for Tview, the website, and server implementation are available at https://git.fmrib.ox.ac.uk/thanayik/dbb. The benefits of Tview are most readily realized with datasets incorporating MRI and microscopy images, enabling visualization of distinct contrasts over multiple spatial scales. However, many datasets provided in the first release to the Digital Brain Bank do not contain any microscopy data. For these datasets, a detailed static image is currently used for visualization on the Digital Brain Bank website. Requirements for data access and referencing datasets The Digital Brain Bank has been designed to minimize the burden on the user to access datasets, within ethical constraints. For many datasets, we have developed conditions of use terms via a material transfer agreement (MTA), which users agree to prior to access. The MTAs are primarily designed to ensure that datasets are used for research/educational purposes, to prevent misuse, and to satisfy funding requirements. For datasets restricted by MTAs, when possible, a subset of example data (e.

Post-mortem magnetic resonance imaging (MRI) provides the opportunity to acquire high-resolution datasets to investigate neuroanatomy and validate the origins of image contrast through microscopy comparisons. We introduce the Digital Brain Bank (open.win.ox.ac.uk/DigitalBrainBank), a data release platform providing open access to curated, multimodal post-mortem neuroimaging datasets. Datasets span three themes-Digital Neuroanatomist: datasets for detailed neuroanatomical investigations; Digital Brain Zoo: datasets for comparative neuroanatomy; and Digital Pathologist: datasets for neuropathology investigations. The first Digital Brain Bank data release includes 21 distinctive whole-brain diffusion MRI datasets for structural connectivity investigations, alongside microscopy and complementary MRI modalities. This includes one of the highest-resolution whole-brain human diffusion MRI datasets ever acquired, whole-brain diffusion MRI in fourteen nonhuman primate species, and one of the largest post-mortem whole-brain cohort imaging studies in neurodegeneration. The Digital Brain Bank is the culmination of our lab's investment into post-mortem MRI methodology and MRI-microscopy analysis techniques. This manuscript provides a detailed overview of our work with post-mortem imaging to date, including the development of diffusion MRI methods to image large post-mortem samples, including whole, human brains. Taken together, the Digital Brain Bank provides cross-scale, cross-species datasets facilitating the incorporation of post-mortem data into neuroimaging studies.

There is a need to understand the histopathological basis of MRI signal characteristics in complex biological matter. Microstructural imaging holds promise for sensitive and specific indicators of the early stages of human neurodegeneration but requires validation against traditional histological markers before it can be reliably applied in the clinical setting. Validation relies on a precise and preferably automatic method to align MRI and histological images of the same tissue, which poses unique challenges compared to more conventional MRI-to-MRI registration. A customisable open-source platform, Tensor Image Registration Library (TIRL) is presented. Based on TIRL, a fully automated pipeline was implemented to align small stained histological images with dissection photographs of corresponding tissue blocks and coronal brain slices, and further with high-resolution (0.5 mm) whole-brain post-mortem MRI data. The pipeline performed three separate deformable registrations to achieve accurate mapping between whole-brain MRI and small-slide histology coordinates. The robustness and accuracy of the individual registration steps were evaluated using both simulated data and real-life images from 6 different anatomical locations of one post-mortem human brain. The automated registration method demonstrated sub-millimetre accuracy in all steps, robustness against tissue damage, and good reproducibility between experiments. The method also outperformed manual landmark-based slice-to-volume registration, also correcting for curvatures in the slicing plane. Due to the customisability of TIRL, the pipeline can be conveniently adapted for other research needs and is therefore suitable for the large-scale comparison of routinely collected histology and MRI data. Highlights * ALS : amyotrophic lateral sclerosis, ANHIR : Automatic Non-rigid Histological Image Registration, BOBYQA : Bound Optimisation by Quadratic Approximation, bSSFP : balanced steady-state free precession sequence, CR : correlation ratio, CT : computed tomography, DOF : degrees of freedom, FSL : FMRIB Software Library, FWHM : full width at half maximum, H&E : haematoxylin and eosin (histological stain), LFB+PAS : Luxol fast blue combined with the periodic acid-Schiff procedure (histological stain), MIND : Modality-Independent Neighbourhood Descriptor, MND : motor neuron disease, MRI : magnetic resonance imaging, NEWUOA : New Unconstrained Optimisation Algorithm, NMI : normalised mutual information, OBB : Oxford Brain Bank, OFC : orbitofrontal cortex, PLP : proteolipid protein, pTDP-43 : phosphorylated TAR-DNA binding protein 43 kDa, SPM : Statistical Parametric Mapping (software), SSD : sum of squared differences, TIRL : Tensor Image Registration Library, TPS : thin-plate spline

We built a novel pipeline to register 290 high-resolution multi-contrast microscopy sections in the BigMac dataset to post-mortem MRI with sub-millimetre accuracy. The MRI-microscopy alignment is optimised over all microscopy sections jointly, to achieve high-quality co-registration whilst ensuring the slide order and separation is consistent with how the tissue was sectioned. We demonstrate the accuracy of the registration by overlaying FreeSurfer-estimated grey-white matter contours on orthogonal sections of 3D volumetric reconstructions of each microscopy modality. Our pipeline provides an integrated workflow for diffusion-microscopy studies using the BigMac dataset, which is openly available via the Oxford Digital Brain Bank.

We present data fusion combining precisely co-registered dMRI and microscopy to reconstruct 3D fibre orientation with micron-scale spatial resolution. We then perform hybrid dMRI-microscopy tractography to investigate two known challenges in tractography: the gyral bias and bottleneck problems. Our results using hybrid tractography suggest that the gyral bias can be overcome by increased spatial resolution, and that microstructure-informed fibre orientations can overcome the bottleneck problem, irrespective of spatial resolution. These observations can be used to inform tractography analyses in vivo. This approach builds on the complementary strengths of microscopy (high spatial resolution) and dMRI (3D whole-brain coverage).

Motivation: Voxel-wise MRI-histology comparisons routinely rely on manual segmentations of ROIs and subjective quantitative histological metrics, an error-prone and labour-intensive process. Goal(s): We developed an automated framework for investigating relationships between multiple MRI metrics and immunostains. Approach: We used MRI and histology protocols optimised for ex-vivo mouse brains. We co-registered this data and derived quantitative histological metrics. We conducted voxel-wise correlations in grey and white matter between MRI (diffusion, R2* and susceptibility) and immunohistochemistry (myelin, neurofilament and extracellular matrix proteins). Results: Our framework successfully recapitulated known relationships for myelin and neurofilaments and, interestingly, demonstrated new relationships between MRI metrics and extracellular matrix protein stains. Impact: Our optimised framework combines openly-shared software and MRI-histology protocols, addressing current challenges, such as obtaining high-quality histology data, MRI-to-histology registration and automatic extraction of quantitative histological metrics. This can benefit future MRI-histology studies in mouse brains prepared for ex-vivo MRI.

Microscopy data can provide detailed descriptions of how cortical microstructure varies across the brain. However, the relationship between cortical microstructure and MRI signals remains relatively unexplored. Here we develop a pipeline to extract cortical profiles from co-registered MRI and microscopy data in the BigMac dataset. We compare cortical profiles of cell density and microstructural dispersion from Nissl-stained histology slides with matched profiles from structural and diffusion MRI. Using CCA we find modes of microstructure variation common to microscopy and structural MRI, but not diffusion, indicating variable sensitivity of the MR metrics to brain-wide variations in Nissl-stained cytoarchitecture.