Sum is more than the parts – the combination of magnetic resonance histology and light sheet microscopy

Light sheet microscopy (LSM) represents a powerful tool for capturing high-resolution images of entire mouse brains, offering minimal photobleaching and phototoxicity. Through the use of diverse immunohistochemical stains, LSM can unveil the distribution and density of distinct cellular populations within the mouse brain. Nonetheless, the accuracy of regional morphological measurements and comparisons with other imaging techniques are hindered by the distortion and tissue tearing incurred during the removal from the skull, fixation, and physical manipulation of LSM data. Additionally, prevailing practices involve aligning LSM data with canonical atlases, such as the Allen Brain Atlas (ABA), which are based on similarly distorted tissue samples. This limitation further undermines accurate regional morphological analysis and cross-modality comparisons. In addition to light sheet microscopy (LSM), magnetic resonance histology (MRH) serves as another valuable technique for high-resolution imaging of mouse brains, offering a range of contrasts such as gradient echo and diffusion tensor imaging. MRH not only provides a reference standard for aligning LSM data and ensuring quality control but also offers supplementary insights into anatomical features and connectivity. However, a drawback of LSM is that the data are typically acquired from ex-skull specimens, leading to inherent distortion and misalignment. Meantime, MRH data lacks the intricate details required to capture the fine-grained cytoarchitecture of the mouse brain.

In a new study published in the journal Frontiers in Neuroscience, researchers Yuqi Tian, James Cook and led by Professor G. Allan Johnson from Department of Radiology at Duke University School of Medicine developed a method to combine light sheet microscopy (LSM) data with magnetic resonance histology (MRH) of the same mouse brain specimen to restore the morphology of the LSM images to the in-skull geometry. This enabled more accurate quantitation of cellular features in specific brain regions and overcome the limitations of existing methods that register LSM data with distorted canonical atlases. Furthermore, the study sought to map condensed set of labels from the common coordinate framework (CCFv3) of the Allen Brain Atlas onto the geometrically corrected full-resolution LSM data.

The research team successfully devised and optimized a registration pipeline capable of aligning large-scale 3D LSM data (reaching terabyte sizes per dataset) with MRH data of the corresponding mouse brain. This pipeline utilized multiple transformation stages at various resolution scales and incorporated a streamlined three-step procedure: pointset initialization, automated ANTs registration with optimized transformation stages, and the application of these transformations to high-resolution LSM data. Remarkably, the registration pipeline achieved accurate alignment within approximately 10 hours, requiring minimal manual intervention. The registered LSM data demonstrated excellent agreement with reference MRH data both at local and global levels. To validate their method, the researchers applied the workflow to a collection of datasets encompassing diverse combinations of MRH contrasts and LSM immunohistochemistry stains. This application yielded a streamlined registration process for LSM images to MRH, facilitating routine analyses. This approach preserved the individual brain morphology and enabled more precise regional annotations and measurements of volumes and cell density.

Overall, the method effectively restored the LSM image morphology to its original in-skull geometry, mitigating the distortion and tissue swelling arising from sample preparation. By aligning LSM data with MRH of the same specimen rather than distorted canonical atlases, the method allowed for more accurate quantitation of cellular features in specific brain regions. The study showcased the feasibility and utility of cross-modality registration between LSM and MRH, underscoring the advantages of MRH as a reference standard for LSM data alignment and quality control. Moreover, the researchers emphasized the complementary information provided by different MRH contrasts and LSM stains. They also suggested future avenues for improvement, such as employing advanced registration algorithms and hardware acceleration to achieve even better results.

In conclusion, Yuqi Tian, and James Cook and Professor Johnson, presented an innovative method that combines LSM data with MRH data from the same mouse brain specimen. This integration successfully restored LSM image morphology to its in-skull geometry and enabled more accurate quantitation of cellular features in specific brain regions. The cross-modality registration between LSM and MRH demonstrated its feasibility and utility, shedding light on the complementary information offered by different imaging techniques. By facilitating the analysis and interpretation of cellular-level features in relation to anatomical structures, this method holds promise for advancing our understanding of the mouse brain. Indeed, the analysis and interpretation of cellular-level features within anatomical structures are indispensable for unraveling the complexities of biological systems, elucidating disease mechanisms, advancing precision medicine, and understanding developmental processes. The new method will stimulate pre-clinical brain research and facilitate many scientific and medical advancements aimed at improving human brain health and well-being.