Imaging intravoxel vessel size distribution in the brain using susceptibility contrast enhanced MRI

Vascular remodeling is inherent to the pathogenesis of many diseases including cancer, neurodegeneration, fibrosis, hypertension, and diabetes. In this paper, a new susceptibility-contrast based MRI approach is established to non-invasively image intravoxel vessel size distribution (VSD), enabling a more comprehensive and quantitative assessment of vascular remodelling. The approach utilizes high-resolution light-sheet fluorescence microscopy (LSFM) images of rodent brain vasculature, gradient echo sampling of free induction decay and spin echo (GESFIDE) MRI signal simulation from the three-dimensional (3D) vascular networks, and training a deep learning (DL) model to predict cerebral blood volume (CBV) and VSD from GESFIDE signals. Specifically, small voxel-size volumes of interest (VOI) (n = 32,000) were extracted from LSFM images of rodent brain and the vascular structure was segmented. Next, two DL models were trained to predict the CBV and VSD from the ratio of pre- and post-contrast GESFIDE signals simulated from these VOIs. The results from ex vivo experiments on test VOIs (n = 3,132) demonstrated strong linear correlation (r = 0.95) and high similarity (mean Bhattacharya Coefficient (BC) = 0.87) between the true and predicted CBV and VSDs, respectively. The DL models outperformed the traditional dictionary-matching approach and demonstrated high accuracy in predicting CBV and VSD, even when the GESFIDE signals were degraded with varying noise levels (SNR: 15, 30, 45, and 60 dB). The DL model showed comparable results to those observed in the test VOIs on a public mouse brain vasculature dataset (n = 1,000), demonstrating the generalizability of the DL models. The accuracy of the predicted CBV (r = 0.78) and VSD (mean BC = 0.82) on the tumor VOIs (n = 706) was moderately high but lower than the accuracy of predicted CBV and VSD observed for the healthy VOIs. Hence, with further in vivo validation, intravoxel VSD imaging could become a transformative preclinical and clinical tool for interrogating disease and treatment-induced vascular remodeling.