Motivation: MRI measurements of microvascular network geometries and functions can help better understand brain functions and pathologies. Goal(s): Improve microvascular network estimates with MR vascular Fingerprinting (MRvF). Approach: We generated 45,000 realistic 3D voxels from mouse brain microscopy datasets using advanced segmentation tools and extracted multiple metrics through graph analysis. Then, we applied data augmentation techniques to generate lesional voxels. Finally, we employed a deep neural network to simulate MR signals and match them to in vivo MR acquisitions on healthy and tumor-bearing animals. Results: Our method improved the MRvF results, allowing the generation of several new microvascular parameter maps. Impact: By combining realistic 3D voxel generation and the MRvF technique, we can achieve a detailed depiction of brain microvascular parameters, potentially improving our understanding of the brain and treatment of brain lesions.
Marçal et al. (Tue,) studied this question.