Diffusion MRI Simulation with the Virtual Imaging Platform

Myocardial fiber architecture plays an important role in ensuring normal mechanical and electrical properties of the heart.However, due to limitations of existing imaging techniques, little is known about the fiber structure of in vivo human heart.Diffusion magnetic resonance imaging (dMRI) is one of the most potential techniques for mapping in-vivo cardiac fiberarchitecture. However, in the absence of the ground truth information and with influence of MRI scanner noise and artifacts, itis difficult to evaluate how well the diffusion characteristics calculated from experimental dMRI reflect the actual cardiac fibermicrostructure properties. To cope with this issue, we propose a dMRI simulator that simulates the diffusion of watermolecules within a virtual cardiac fiber model with a Monte-Carlo approach, and calculates the diffusion images using thespin quantum theory. The noise-free and non-artifacts images from this simulation enable the evaluation of dMRI imageprocessing algorithms and the optimization of imaging parameters based on the model ground truth. To support its computing needs, the simulator was ported to the European Grid Infrastructure (EGI) through the VirtualImaging Platform (VIP). The simulation of 16 billion water molecules, which would require 8 CPU years on a representativemachine of the biomed virtual organization (VO), can now be achieved in one month. Much manual intervention is stillrequired to compute such simulations though