Method of Hypotheses Proving Based on Statistical Processing of Heterogeneous EEG Data

This article describes the automatization of the hypothesis proving with heterogeneous EEG data obtained using different equipment (Neuroscan and Brain Products). EEG data is recorded from the head surface of the subjects by a helmet, which has 118–128 electrodes (channels). The subject undergoes various tests (trials) during the experiment The EEG data is divided into parts, in accordance with the trials. The connection between electrodes and head is unreliable, some channels may be lost because of the signal weakness. The approach includes unification of EEGrecordings,
restoring of lost channels, 3D reconstruction of brain activity and conducting a mediation analysis. Raw data is overwritten in a unified order. Unmatched channels are excluded. Single corrupted channels are restored by spherical spline interpolation. 3D localization of brain activity is based on the inverse problem. The localization is conducted in the functional areas of the cerebral cortex, according to the Talayrak atlas using the MN method. Reconstruction is carried out separately for each trial, and for each of the five standard frequency ranges. The results are recorded in the NIFTI format, focused on the voxel representation. Then a multi-level mediation analysis is carried out. The coordinates of the discovered clusters are compared to the brain map and serve as a basis for interpreting and verifying neurophysiological hypotheses. The approach was implemented as a set of MATLAB scripts, libraries: EEGLAB, NeuroElf, Alphasim, spm8, Mediation toolbox, and the sLORETA software package. The created toolshave been practically tested in the processing of neurophysiological experiments on social interaction. Created scripts can be used to test a wide class of neurophysiological hypotheses.

Keywords

hypothesis proving, electroencephalography, data processing, 3D reconstruction, mediation analysis

Acknowledgments
The study was supported by the Russian Science Foundation (RSF) under Grant № 17-18-01019