The Bayesian source estimation method “Sequential Semi-Analytic Monte-Carlo Estimation”, or in short SESAME, allows a largely un-biased dipole modelling approach to the inverse problem of reconstructing brain activity from the few measurement points that an M/EEG sensor array provides. It is based not on minimization of residuals, energies, or other deterministic rules, but instead on a probabilistic formulation of the problem, and yields the most likely configuration that explains the data.
In the past, the success of the method was somewhat hampered by the difficulty of estimating the probability of source strengths. In the recently published article “Where Bayes tweaks Gauss: conditionally Gaussian priors for stable multi-dipole estimation” by Viani et al. in Inverse Problems and Imaging (doi: 10.3934/ipi.2021030), the authors describe how introducing a hyperprior effectively removes the dependency of the solution on the choice of this parameter, thus yielding extremely robust estimates. The overall effect of this is a further reduction in user-bias, along with a further improvement in the reliability of the SESAME results. This improved approach is available in BESA Research 7.1.