Data analysis¶
the data we will analyse was downloaded from https://repod.icm.edu.pl/dataset.xhtml?persistentId=doi:10.18150/repod.0107441
we want to use MNE-Python for the analysis
we want to look at the data of all subjects to identify ‘bad channels’ and artifacts in the signals
Overview¶
first, we want to do some preprocessing with the eeg data, means filtering, bad channel handling (interpolation or not (?)) and artifact correction
afterwards, we want to do some statistical analyses with e. g. panda (python based)
goal could be to find some significant differences in eeg signals between schizophrenia patients and healthy controls
Outcomes¶
A general overview about the analysis steps. For a closer look, please look at our codes.
Preprocessing
imported modules:
numpy
mne
os
set directory to data folder to load the .edf files
check up the raw data
creating a template for excluding eye artifacts
build up ICA (seven components, save these files as .fif files)
Frequency-Band Analysis
import mne and numpy module
load average for each frequency activation for every person and define minimum/ maximum power frame
load all filtered and preprocessed eeg/ .fif files for every person
define the files referring to one of the two groups (schizophrenic and healthy persons)
create epochs
averaging the activity to do some statistics
plot topomaps for all frequencies for the two groups
statistics
import seaborn, pandas and matplotlib
detect which electrodes might be of critical interest in comparison between healthies and schizophrenics for each frequency activation
define critical electrodes for each frequency acitvation
load and show swarm and box plots for average activation for healthies and schizophrenics
import scipy
do independent t-test for the two variables to find out if there is a significant difference