Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 10 (1), 5504

Multimodal Image Registration and Connectivity Analysis for Integration of Connectomic Data From Microscopy to MRI

Affiliations

Multimodal Image Registration and Connectivity Analysis for Integration of Connectomic Data From Microscopy to MRI

Maged Goubran et al. Nat Commun.

Abstract

3D histology, slice-based connectivity atlases, and diffusion MRI are common techniques to map brain wiring. While there are many modality-specific tools to process these data, there is a lack of integration across modalities. We develop an automated resource that combines histologically cleared volumes with connectivity atlases and MRI, enabling the analysis of histological features across multiple fiber tracts and networks, and their correlation with in-vivo biomarkers. We apply our pipeline in a murine stroke model, demonstrating not only strong correspondence between MRI abnormalities and CLARITY-tissue staining, but also uncovering acute cellular effects in areas connected to the ischemic core. We provide improved maps of connectivity by quantifying projection terminals from CLARITY viral injections, and integrate diffusion MRI with CLARITY viral tracing to compare connectivity maps across scales. Finally, we demonstrate tract-level histological changes of stroke through this multimodal integration. This resource can propel investigations of network alterations underlying neurological disorders.

Conflict of interest statement

The authors declare no competing interests.

Figures

Fig. 1
Fig. 1
MIRACL enables the interrogation of brain pathways and cellular features across modalities. The resource integrates CLARITY data in the microscopic domain (top left) with macroscopic in vivo and ex vivo imaging data, such as structural MR, diffusion MR, and quantitative MR relaxometry (top right), in the Allen atlas reference frame “ARA” (top center). This integration enables tract and network-based analyses such as studies of histological features across network graphs or along fiber tracts and connectivity analyses based on projection terminals. This pipeline also performs group-level statistics, multimodal correlations, as well as comparisons of connectivity maps across scales.
Fig. 2
Fig. 2
Validation highlighting MIRACL’s registration and segmentation accuracy. a High fidelity of CLARITY registration to the Allen Regional Atlas (ARA). Left: Coronal view of an ARA Nissl histology slice. Center: CLARITY auto-fluorescence channel of a representative stroke mouse registered to the ARA. Right: The same auto-fluorescence channel with overlaid label outlines (R–L: right–left, S-I: superior–inferior). b Axial and coronal views of a Thy1 yellow fluorescence protein (YFP) imaging volume (green) registered to the ARA template (grayscale) and an axial view of another registered CLARITY dataset with (right) and without (left) ARA labels. c Coronal views of three Allen-registered in vivo MR images (a control mouse, a striatal stroke mouse, and a cortico-striatal stroke, respectively). All mice in this study were scanned 24 h after stroke. d Low root mean squared error (RMSE) between transformed manually placed landmarks on the native MRI and CLARITY imaging volumes and ARA manually placed landmarks. Center line of box plot represents the median, bounds represent the first and third quantiles, and whiskers represent the lowest and highest datum within 1.5× the interquartile range of the lower and upper quantiles. e Segmentation results for nuclei using propidium iodide (PI) stain. Coronal view of a PI stroke brain and its corresponding segmentation image (scale bar: 400 µm). Inset (right) shows a zoom-in view on cortex ipsilateral to the stroke with individually segmented cells shown in random colors overlaid on the original PI image (scale bar: 100 µm). f Segmentation results for layer-specific neurons using Thy1-YFP (YFP). Coronal view of a Thy1-YFP stroke brain and its segmentation image (scale bar: 400 µm). Insets (right) show zoom-in views on the cortex contralateral to the stroke (orange box) and cortex ipsilateral to the stroke (blue box) (scale bars: 50 µm). g Zoom-in on YFP results with segmentation overlaid on raw images, and 3D rendering of YFP raw and segmented neurons. h 3D rendering of an original YFP volume with a 5-µm isotropic resolution (left) and examples of voxelized segmentation results (where the segmentation images are summarized at lower resolutions in the Allen space) at 25 µm in two- and three-dimensions (scale bar: 600 µm).
Fig. 3
Fig. 3
MRI measures correlate with CLARITY features in acute stroke. a Voxel-wise heat-maps (averaged across mice) depict the relative levels of yellow fluorescent protein (YFP), propidium iodide (PI), transverse relaxation time (T2), and mean diffusivity (MD) averaged over all 10 stroke mice, demonstrating acute changes in right-hemisphere stroke discerned by both CLARITY and MRI (white arrows). b Group-level statistical maps showing differences between ipsilateral and contralateral regions for Thy1-YFP, PI, T2, and MD. Color bar represents significance levels (label-wise p values) thresholded at alpha 0.05. c CLARITY–MRI correlations demonstrating a positive correlation of MD and a negative correlation of T1 with neuronal counts. Voxel-wise heat-map correlations of PI cell counts with quantitative MRI parameters (relaxometry: T2/T1, diffusion: MD, fractional anisotropy (FA)). Spearman’s rank correlations between PI cell counts and MRI parameters show a substantial decrease in MD values as well as an increase in quantitative T2 and T1 values in the stroke region (red) as compared to the unaffected brain regions (blue).
Fig. 4
Fig. 4
Observed cellular degeneration in remote region at the acute stage. a Visualization of cellular degeneration (relative to controls) due to stroke in ARA labels. From left to right: Binary PI segmentation of a control mouse (coronal view) showing similar nuclei counts across both hemispheres. Binary PI segmentation of a stroke mouse showing diminished cellularity in the infarct. The voxelized map (in blue, downsampled and registered to the ARA template in grayscale) also shows diminished cell density per voxel. Right: Two coronal slices of the ARA atlas with label intensity representing decreased cell density normalized to control mice. The dotted line approximately outlines the stroke region with a 50% incidence across the stroke cohort. Lower opacity corresponds to a decrease in cell density. SS: somatosensory areas, CP: caudoputamen, TH: thalamus, HP: hippocampus, CEA: central amygdalar nucleus, BLA: basolateral amygdalar nucleus, S–I: superior–inferior, R–L: right–left. b Investigation of cellular changes in the connected regions. Left: stroke incidence map demonstrating the incidence of an ARA voxel being present in the stroke cohort, with the black dashed line outlining an incidence of >50%. Center left: Rendering of the 50% stroke incidence mask depicting the largest five ARA labels in that mask (RCP: right caudoputamen, RPIR: right piriform area, RCA1: field CA1 of the right hippocampus, RCA3: field CA3 of the right hippocampus, RSSp-bfd: right primary somatosensory area, barrel field). Center right: projection targets of each of the largest five labels in the stroke mask (with each label as an injection sites), ranked by normalized projection volume (color bar). Right: The bar graphs represent PI cell density in our stroke mice normalized to controls. Salmon-colored bars indicate regions with cell densities lower than two standard deviations from the mean of all normalized regions. The strength of connectivity decreases from left to right and with increasing color brightness, with the most connected regions being darker and on the left. Individual data points for each target region are overlaid on its corresponding bar graph. Error bars represent 95% confidence intervals using 1000 bootstrap iterations.
Fig. 5
Fig. 5
Examination of the involved networks and connected hubs in the stroke region. a Projection density map of the five largest regions in the stroke incidence mask (dashed line in Fig. 4b) as primary injection sites (y-axis) and their top five connectivity targets, ranked by connectivity strength (x-axis). Cell color in both the left and middle panels represents normalized projection volume between the primary injection (y-axis) and target structure (x-axis). b Matrix of the largest ten labels, as primary injection sites (y-axis), in the stroke area and their ten targets most common across all injection sites. c Snapshot of an interactive network graph of the largest 25 labels in the stroke mask and their 25 most common targets, highlighting example hubs (regions with a high degree of connectivity) inside and outside the stroke region. Blue nodes (circles representing brain regions) are regions inside the stroke and purple nodes are outside. Example hubs within the stroke: CP: caudoputamen, PIR: piriform area, SS: somatosensory areas.
Fig. 6
Fig. 6
Computing streamline terminal zones and connectivity graphs from CLARITY viral tracing. a Generation of the structure tensor analysis (STA)-based tract density and remote tract terminal maps. First panel: Anatomical locations of several nodes of the medial prefrontal cortex (mPFC) network. Second panel: Maximum intensity projection of an original CLARITY volume with an AAV tracing injection at the mPFC. Third panel: Tract density maps (red) of STA-based streamlines from a CLARITY volume registered to the ARA. Fourth panel: Tract terminals map (blue dots) from the same specimen, both overlaid over the 25 µm ARA template and tract density map (red). Fifth panel: The same tract terminals map (blue dots) representing only terminals of long-range mPFC streamlines, with terminals near the injection site masked out, overlaid on the tract density map (red). b STA-based tractography of CLARITY viral tracing, colored by terminal location. c Depicting passing versus terminating mPFC tracts. mPFC streamlines to the ventral tegmental area (VTA), nucleus accumbens (ACB), and anterior cingulate area (ACA). Magnifications of passing versus terminating fibers in the VTA and ACA show more fibers terminate in the VTA than in the ACA.
Fig. 7
Fig. 7
Examining prefrontal fiber projections using dMRI, the Allen connectivity atlas, and CLARITY viral tracing. a Raw and processed data across different connectivity scales (axial and sagittal views). Left panel: Ex vivo diffusion tensors (primary eigenvector for each voxel) overlaid on a fractional anisotropy (FA) map. Center left panel: Fiber orientation distribution computed using a constrained spherical deconvolution model (CSD, ref. ) and overlaid on the FA map. Center right panel: Projection density map from a medial prefrontal cortex (mPFC) injection from the Allen connectivity atlas (red), overlaid on the 3D ARA template (grayscale). Right panel: Raw images from an example CLARITY viral tracing experiment with an mPFC injection, overlaid on a slice from the ARA template. A–P: anterior–posterior, R–L: right–left. b Tractography and fiber segmentation across different modalities from a sagittal view. The first two images show fiber tractography streamlines from diffusion MRI with different tracking techniques (deterministic tensor and probabilistic CSD), followed by a rendering of Allen connectivity, and lastly structural tensor analysis (STA) reconstructed streamlines from CLARITY in a. ACB: nucleus accumbens, CST: corticospinal tract, PL: prelimbic area of mPFC, VTA: ventral tegmental area. S–I: superior–inferior, R–L: right–left. c Resolving branching of fibers projecting from the prelimbic area (PL) to the mediodorsal nucleus of the thalamus (yellow arrow). Deterministic tensor tractography fails to capture this fiber branching, while probabilistic CSD tractography reproduced this projection that was confirmed in the Allen atlas and our CLARITY viral tracing data.
Fig. 8
Fig. 8
Thy1-YFP cellular modulations along dMRI-based tractography in acute stroke. a Thy1-YFP CLARITY axial images from a stroke mouse showing lower staining on the right hemisphere (right caudoputamen outlined). Images were intensity corrected for this figure (not for the analysis) to account for signal drop at the base of the olfactory bulb. b Two tracts were identified by probabilistic CSD tractography on our ultra-high resolution ex vivo dMRI dataset, seeding (starting from) the mPFC and targeting either VTA or RSP. Tracts were filtered to include only terminating fibers. c Cellular modulation along ipsilateral VTA but not RSP tracts in stroke mice compared to controls. Left: Track-based Thy1-YFP asymmetry depicted along these two tracts. Color scale represents percentage of asymmetry. Center: Tract-based Thy1-YFP intensity for the two tracts ipsilateral (red) and contralateral (blue) to the stroke, showing a decrease along the tract targeting the VTA but not the RSP. Right: Mean tract-level difference between contralateral and ipsilateral Thy1-YFP intensity, divided by the mean of contralateral Thy1-YFP intensity. x-axis of the graphs represents nodes (locations) along tracts going anterior (left) to posterior (right). Center line of box plot represents the median, bounds represent the first and third quantiles, and whiskers represent the lowest and highest datum within 1.5× the interquartile range of the lower and upper quantiles.

Similar articles

See all similar articles

References

    1. Denk W, Briggman KL, Helmstaedter M. Structural neurobiology: missing link to a mechanistic understanding of neural computation. Nat. Rev. Neurosci. 2012;13:351–358. doi: 10.1038/nrn3169. - DOI - PubMed
    1. Silasi G, Murphy TH. Stroke and the connectome: how connectivity guides therapeutic intervention. Neuron. 2014;83:1354–1368. doi: 10.1016/j.neuron.2014.08.052. - DOI - PubMed
    1. Zhang JH, et al. The vascular neural network—a new paradigm in stroke pathophysiology. Nat. Rev. Neurol. 2012;8:711–716. doi: 10.1038/nrneurol.2012.210. - DOI - PMC - PubMed
    1. Weishaupt N, Zhang A, Deziel RA, Tasker RA, Whitehead SN. Prefrontal ischemia in the rat leads to secondary damage and inflammation in remote gray and white matter regions. Front. Neurosci. 2016;10:81. doi: 10.3389/fnins.2016.00081. - DOI - PMC - PubMed
    1. Sozmen EG, Kolekar A, Havton LA, Thomas Carmichael S. A white matter stroke model in the mouse: Axonal damage, progenitor responses and MRI correlates. J. Neurosci. Methods. 2009;180:261–272. doi: 10.1016/j.jneumeth.2009.03.017. - DOI - PMC - PubMed

Publication types

Feedback