Diffusional kurtosis imaging in brain MRI using biexponential model

Abstract

Diffusion-weighted imaging (DWI) is a method that uses specific Magnetic Resonance Imaging sequences, which make signal decrease where diffusion occurs. Conventional diffusion signal model is mono-exponential model which assumes that there is the only one type of diffusion in a voxel. However, in human body, the diffusion of proton does not follow Gaussian distribution due to the complex structure of biological tissues, such as membrane, fiber, etc. In this reason, mono-exponential model is not sufficient to fully describe diffusion in human body.

Bi-exponential model has been proposed to explain the intra-cellular diffusion and extra-cellular diffusion. Fast diffusion coefficient, slow diffusion coefficient, and fraction are calculated by fitting diffusion weighted signals. Besides bi-exponential model, kurtosis imaging which is a measure of how deviates from Gaussian distribution is also used widely to examine diffusion in human body.

Proposed algorithm provides the information of kurtosis of slow diffusion. Fitting algorithm was validated by simulation.