paperKB
coga / coga-kb
Processing
Help
Sign in

Chunk #10 — Materials and methods — Feature extraction — EEG extracted features

Source
Predicting alcohol use disorder remission: a longitudinal multimodal multi-featured machine learning approach.
Embedded
yes

Text

A full description of EEG functional connectivity calculation (using MNE package)24 can be found in Supplementary Materials and Supplementary Table S5. Briefly, The FreeSurfer parcellation scheme (aparc.lh/rh), based on the Desikan–Killiany Atlas25, was used to define 68 cortical regions from both hemispheres (list of ROIs in Supplementary Table S5). We computed spectral coherence26 to measure functional connectivity (FC) between EEG signals of 68 regions of interest (ROI) at specific frequency bands: theta (4–8 Hz), alpha (8–12 Hz), beta (12–30 Hz), and gamma (30–60 Hz) with no overlap between frequencies. The following electrophysiological features were extracted: for each of the frequency bands (theta, alpha, beta, and gamma), a 68 × 68 ROIs matrix of coherence was created for each participant resulting in 9221 features. Each of these features represents an EEG coherence functional connectivity (EEG-FC) between two ROIs.