A conditional autoregressive model for estimating slow and fast diffusion from magnetic resonance images

Abstract

The Intra-Voxel Incoherent Motion (IVIM) model is largely adopted to estimate slow and fast diffusion parameters of water molecules in biological tissues, which are used as biomarkers for different diseases. However, the standard approach to obtain the maps of these parameters is based on a voxel-by-voxel estimation and neglects the spatial correlations, thus resulting in noisy maps. To get better maps, we propose a Bayesian approach that exploits a Conditional Autoregressive (CAR) prior density. We consider a pure CAR model and a mixture CAR model, and we compare the outcomes with two benchmark approaches. Results show better maps under the CAR models.