Diffusion kurtosis imaging: Monte Carlo simulation of diffusion processes using crowdprocess

Abstract

Diffusion Kurtosis Imaging quantifies the non-gaussianity of water diffusion. The sensitivity of the diffusion kurtosis model to the microstructural heterogeneity of biological tissues can be studied using Monte-Carlo simulations, which being computationally expensive, can be better applied using parallel computing. To demonstrate the advantages of simulating diffusion processes in distributed computing, in this study, we focused on the correlation between kurtosis of diffusion and three important microstructural changes: intracellular volume fraction, cell radii and cell permeability. A comparison was made between processing in a common laptop CPU, and in CrowdProcess, a new method of parallel computing that uses a browser-powered distributed computing platform. CrowdProcess has proved to be extremely fast compared to the traditional method, bringing the possibility of increasing simulations numerical precision without the time-consuming cost.