Fig. 3: Analysis of simulated EEG signals and low-dimensional features extraction.

Panel A: the simulated data are reconstructed at the sensor level for the channels in the motor frontal area, using the lead-field matrix. Panel B: the process of feature extraction from the z-score normalized EEG synthetic data, with the threshold set to θ = 1.5. Panel C: variations of the simulated features as a function of varying w_dopa. The range of colors signifies the increase in w_dopa. Panel D: validation of the SBI pipeline on 5 random synthetic datasets uniformly selected with w_dopa drawn from an interval ranging from 0.05 to 5. The left plot illustrates the estimated values as a black dashed line, corresponding to \({\tilde{w}}_{{\rm{dopa}}}\), which are interpolated by a linear regression line closely following the red dashed line, indicating a perfect fit. The right plot displays the distribution of the posterior z-score versus the posterior shrinkage, indicating an ideal Bayesian estimation (corresponding to z-scores close to zero and shrinkages close to one).