*2.6. Source Localization*

We applied low-resolution precision electromagnetic brain tomography (eLORETA) to solve the inverse problem and localize the sources of neuronal activity according to EEG data at each of the predetermined points (voxels) in the brain volume [22–24].

LORETA is low-resolution brain electromagnetic tomography. This method solves the inverse problem: converting EEG measurements into information about the distribution of neural sources power into a brain volume. This method belongs to the class of nonparametric methods [25], which are based on the assumption that a separate current dipole (a source) is assigned to each of tens of thousands of elements of the tessellation of the cerebral cortex, while the orientation of the dipole is determined by the local normal to the surface. In this case, the inverse problem is linear, since the only unknowns are the amplitudes of the dipoles. Exact low-resolution brain electromagnetic tomography (eLORETA) is 3D, regularized, and minimum norm-weighted inverse solution with theoretically accurate zero error localization, even in the presence of structured biological or measurement noise [22,25]. The "Colin27" brain MRI averaged template [26] was used to develop a three-layer (brain, skull, and scalp) head model based on a boundary element method (BEM) [27,28]. The sources space inside the brain consisted of 11,865 voxels. The location of the EEG electrodes corresponded to the template head shape.

We analyzed the source characteristics in the predefined time-frequency domain of interest, selected on the basis of sensor-level analysis. To do this, we reassigned the EEG signals to the total average, subtracted the mean, and filtered with a fourth-order Butterworth [ *fL*, *fH*]-Hz band-pass filter, where *fL* and *fH* define the frequency domain of interest. Then, we performed time-lock averaging across the TOI1 trials and computed the covariance matrix. The inverse solution yielded estimates of the source power in each voxel, averaged over the selected TOI window. Finally, we normalized the obtained estimates of the power *P* of each source to the power of 40-s EEG resting state EEG as (*P* − *Prest*)/*Prest*. We used the automated anatomical labeling (AAL) brain atlas [29] to map the location of sources to the anatomical brain regions.
