Quantitative BOLD with Variational Bayesian inference: model comparisons with Monte Carlo simulations and in an elderly cohort

This study uses Monte Carlo simulations to understand the behavior of quantitative BOLD in different physiological conditions, using a Variational Bayesian inference framework with prior information and data from Asymmetric Spin Echo (ASE) scans. The performance of the three models at 7 SNR levels (from 5 to 500) showed that one-compartment and two-compartment models estimated oxygen extraction fraction (OEF) more accurately than linear model across a full range of deoxygenated blood volume (DBV) (p<0.05 using two-way ANOVA with pairwise comparisons). In vivo data showed that Bayesian inference approach effectively enables smoother and quantitatively different parameter maps of OEF and DBV.