Does Kurtosis or Stretched-Exponential Model Fit Experimental Diffusion-Weighted Data Better?

Abstract

Diffusion-weighted (DW) signal dependence on b-value in neural tissues deviates from monoexponential decay. As a result, numerous models have been proposed to characterize such non-monoexponential decay. Two of them, namely kurtosis and stretched-exponential model, are of special interest. This study aimed to evaluate the performance of these two models in describing the experimental DW signals obtained from rodent brains in vivo. Our analysis demonstrates that the quadratic diffusion kurtosis model provides better fitting of DWI data, suggesting that DKI is a more accurate and robust diffusion model for characterizing the complex diffusion processes in vivo in neural tissue.