Where Bayes tweaks Gauss: Conditionally Gaussian priors for stable multi-dipole estimation

Figure 1.  An example of simulated EEG (top) and MEG (bottom) recordings

Figure 2.  Confusion matrices for the estimated number of dipoles, for three different values of the prior scale factor $ k $: in the top panel results obtained with simulated EEG data, in the bottom panel results obtained with simulated MEG data

Figure 3.  Boxplots of the OSPA metric, quantifying the distance between the true and estimated dipole configurations, for three different values of the prior scale factor $ k $: in red the SESAME results, in blue the h-SESAME results; in the top panel results obtained with simulated EEG data, in the bottom panel results obtained with simulated MEG data

Figure 4.  Variance of the posterior probability map with respect to different values of the prior scale factor $ k $, as defined in eq. (16): in red the SESAME results, in blue the h-SESAME results

Figure 5.  Estimated value of the prior width $ \sigma_q $, for different values of the prior scale factor $ k $

Figure 6.  Experimental MEG data: averaged response to auditory stimuli; data taken from the sample open dataset within the MNE–Python package

Figure 7.  Posterior probability maps (left) and estimated source time courses (right) obtained by SESAME when applied to the experimental data shown in Figure 6 with different values of the prior scale factor: $ k = 0.1 $ (top row), $ k = 1 $ (middle row) and $ k = 10 $ (bottom row)

Figure 8.  Posterior probability maps (left) and estimated source time courses (right) obtained by h-SESAME when applied to the experimental data shown in Figure 6 with different values of the prior scale factor: $ k = 0.1 $ (top row), $ k = 1 $ (middle row) and $ k = 10 $ (bottom row)