BRAIN: 3D CHARACTERIZATION OF BRAIN VASCULATURE AND MICROGLIA TO UNCOVER MORPHO-PHYSIOLOGICAL ALTERATIONS IN ALZHEIMER’S DISEASE MURINE MODELS

The brain is a complex organ with a finely tuned organization among its components. Neurological disorders cause structural and physiological alterations not only in neurons but in key elements such as vasculature and microglia; yet understanding their multiscale morphology and topology remains challenging. Building on recent advances in three-dimensional (3D) imaging and spatial transcriptomics, our project, brAIn, aims to understand microglia-vasculature interactions in health and disease using artificial intelligence (AI) for 3D image analysis. To perform 3D quantifications, we analyze large confocal volumetric images of clarified mouse brains from wild-type and Alzheimer’s disease (AD) mouse models labeled for microglia and vasculature. These images, together with paired annotations, are needed to train a ResUNet model for automatic segmentation of both structures. Although our laboratory has already trained a deep-learning model for microglial cell-body segmentation, manual annotation of vasculature remains complex and time-consuming; accordingly, we have employed a generative AI-based strategy to create a synthetic training dataset. Generative adversarial networks combined with fractal-based modelling (L-Systems) are used to mimic the style and properties of the original images within binary-modelled vasculature. These models allow the extraction of 3D morphological and organizational information from microglia and vasculature to study their impact on health and disease, including the presence of Aβ plaques. Finally, imaging-derived features will be integrated with MERFISH single-cell spatial transcriptomics to unravel their molecular interactions within diverse neural tissues. Ultimately, our work aims to elucidate novel mechanisms underlying microglia-vasculature interactions, improving our understanding of brain disease progression and its morphological implications.