VASAL: A Vascular Substrate Algorithm to Generate Microvascular Network Phantoms

To date, few methods have been proposed for synthesising realistic microvascular networks, particularly with compatibility for diffusion MRI (dMRI) simulations. This work presents a generative algorithm for building computational brain microvascular network phantoms, together with a Monte Carlo framework for simulating spin flow dynamics and corresponding dMRI signals. Phantom morphology can be user-tuned according to the: number of generations (branches); total number of segments; segment diameter and length; angle between connecting segments, and; maximum extent in space. The geometric fidelity of resulting VASAL phantoms was validated against an exemplar rat hippocampal network. Comparable distributions of segment lengths, diameters and orientations, and branching characteristics were obtained, demonstrating VASAL replicated realistic microvascular structures. Implementation of spin flow dynamics was validated by simulating dMRI signals in the ballistic regime, which has a well-defined analytical solution; agreement was excellent between simulated signals and analytical solutions (average root mean square error of 0.005). Following this validation of network geometry and spin flow dynamics, VASAL phantoms with different morphologies were generated to evaluate key IVIM model assumptions. Phantom geometry substantially affected dMRI signals and caused deviations from the IVIM model: for example, complex branching phantoms displayed non-Gaussian signal attenuation and diffusion time dependence. While the IVIM model was exemplified here, the presented VASAL algorithm holds potential for advancing other dMRI acquisitions and models: by enabling simulation of complex blood flow within physiologically accurate vascular networks, a route to disentangling how vascular structure and function influence dMRI signals is initiated.