Improving microstructural estimation in time-dependent diffusion MRI model with a Bayesian method

In this study, we proposed Bayesian estimation of tissue microstructures in td-dMRI model, and compared its performance with the traditional non-linear least square fitting method in simulation data and glioma patient data. We found that the performance of Bayesian fitting was dependent on the prior distribution and choice of hyperparameters, and the combination of Bayesian and least-square fitting could achieve reasonable performance without prior information.