Voxel to voxel validation of brainstem MRI novel sequences to postmortem tissue reconstruction

Abstract Background Neuroimaging of brainstem structures has gained importance in ADRD as these regions degenerate early in AD. In recent years there has been an explosion of novel neuroimaging sequences to image the brainstem. However, most of these methods lack validation against the ground truth, i.e tissue visualization, making these methods less attractive for clinical use. Among the new proposed methods, high‐definition connectome promises to revolutionize BS imaging as it overcomes several limitations of standard structural MRI, including a high level of imaging deformation. In this project we have developed a pipeline to validate BS connectome sequences using a voxel to voxel comparison between 3D histological volumes and connectome signal. To this date we applied this pipeline to 6 human brains. Method We used 6 AD cases collected by the Neuroradiology Lab at the University of Sao Paulo. Patients underwent postmortem MRI prior to autopsy.We used a pipeline developed in‐house that allows for precise alignment of histology‐based cytoarchitectural 3D reconstructions of the brainstem to its T2 weighted 7T‐MRI counterpart. Upon procurement, the brains were fixed in formalin, embedded in celloidin, and cut in horizontal sections of 300um. Selected BS structures were delineated on these slides stained for Nissl(gallocyanin)(Fig.1). Delineated images underwent manual quality checks, a series of alignment and registration processes to recover original coordinates. The stained sections and corresponding ROI masks were registered to the blockface to correct for any deformation of tissue during the staining process. Freeview commands applied to reconstruct the brainstems and masks. We registered the resulting 3D reconstruction to the matching MRI coordinates(Fig. 2), enabling voxel‐to‐voxel comparisons. Result 6 cases went through the pipeline with ROIs reconstructed on block reconstructions(Fig. 3). Currently, dice coefficient analysis between MRI signal and the manually outlined ROIs on histology are being finalized. Conclusion Our pipeline enables voxel‐to‐voxel correlation analysis between histology and MRI scans for validation of brainstem neuroimaging. Current finalization of a total 9 cases is ongoing. Validation of such correlation maps will inform interpretations of connectome results in living patients with the potential to act as an eventual diagnostic tool for ADRD.