Estimation of relaxation time distributions in magnetic resonance imaging

Recently there has been increasing interest in estimating the distribution of relaxation times contributing to a magnetic resonance signal. This paper shows that it is impractical to estimate the spread of such a distribution from typical measurements. Instead, a Bayesian estimator is developed for a discrete distribution, which is very robust to noise. Although the distribution spread is not modelled, the estimates capture the main features of the distribution such as the mode locations and often provide improved myelin water fraction estimates in simulation examples.