Source Localization

Source localization of brain-related activity on the surface of the cortex was determined using LORETA (Low-Resolution brain Electromagnetic Tomography) Key software employing sLORETA (Fuchs et al., 2002; Pascual-Marqui, 2002; Jurcak et al., 2007). LORETA is a low-resolution source localization technique that computes the generated electric neural activity in voxels that segment the gray matter of the brain. Only the ICs determined to be brain components by ICLabel were used. All other components containing artifacts were removed from the EEG data. The Electrode position on the head was determined by reference to the 10-10 system within the LORETA Key software. It should be noted that because generic head models were used instead of individual-specific models based on MRI anatomical images in which the position of the electrodes in relation to the brain can be precisely determined the absolute accuracy of the sLORETA source localization is limited, although localization is accurate at a regional level of resolution.

Source localization was computed for 21 improv and 21 scale trials in total, in the following frequency bands (theta 6.5–8 Hz; alpha1 8.5–10 Hz; alpha2 10.5–12 Hz; beta1 12.5–18 Hz; beta2 18.5–21 Hz; beta3 21.5–30 Hz; gamma 30.5–50 Hz; full alpha 8.5–12 Hz; full beta 12.5–30 Hz). The frequency bands were selected in part because they were defaults given in the Loreta Key software that are consistent with those given in Rangaswamy et al. (2002), Kropotov (2016), and Newson and Thiagarajan (2019). Another practical reason for using multiple sub-bands within traditional frequency ranges was to increase the number of features for machine learning analysis (see below). For each participant the individual trials were submitted to a statistical non-parametric (SnPM) analysis for the contrast of improv vs. scale. The resultant sLORETA map for this contrast for each participant was used in a random-effects group-level analysis using SnPM analysis (5,000 randomizations) within the LORETA Key software. Corrected critical thresholds for multiple comparisons were used in accordance with SnPM (Nichols and Holmes, 2001).

We believe that there are advantages in the analysis of differences between improv and scale at the cortical source localized level (using sLORETA) over that of conducting the analysis over the separate brain-related ICs. In this research our goal was to use ICA to remove artifacts from the EEG while retaining brain-related activity. The brain-related components were determined by ICLabel and then all other components were removed from the EEG data. Although it is possible to do spectral analysis of the IC activations contrasting improv vs. scale conditions for a single participant it is often difficult to find comparable components that are the same across all participants. One limitation of the cluster analyses used to group similar ICs across participants is that they can give variable results depending on the number of clusters the algorithms are set to find. An advantage of utilizing all brain-related ICs for the EEG data is that their mixture may result in differences in cortical source localized activity that may not be found when considering the individual ICs separately.