Preprocessing pipeline for EEG Data using EEGLAB.

Preprocessing pipeline for EEG Data using EEGLAB. For data collected using BrainVision and EGI/Philips systems.

The preprocessing script performs the following steps: -

1. Read data
2. Downsample to 250 Hz
3. Filter using 0.3 to 20 Hz band-pass filter
4. Offline rereference to average of linked mastoids
5. Remove channels (say in the periphery) that are not needed for analysis
6. Remove bad channels; these will be interpolated later in step 9
   1. This will be referred to as original data below
7. Run Indepedendent Components Analysis (ICA)
   1. Make a copy of the original data from step 6 and name it as EEG_ICA
   2. Downsample EEG_ICA to 100 Hz
   3. Filter EEG_ICA using 1 Hz high-pass filter (i.e., 1 Hz low cut off)
   4. Delete first 60 s and last 30 s of the recording from EEG_ICA
   5. Run extended Infomax ICA
   6. Run ICLabel and flag artefactual ICA components for rejection
   7. Copy ICA weights and related information from EEG_ICA to original data
   8. Calculate the ICA activation matrix in the original data that now contains ICA information
   9. Copy the result of ICLabel and component rejection markings from EEG_ICA to the original data
   10. The original data now has ICA information, including components marked as artefactual
8. From the original data from step 7.10, remove the components marked as artefactual
9. Interpolate bad channels, if any, that were removed in step 6
10. Save the ICA pruned original data for further analysis

Notes

1. No continuous data rejections / epoch rejections are performed. Therefore no new boundary events are introduced to the data, and therefore the overall length (i.e., duration in time) of the pruned data for analysis should be identical to that of the raw data.
2. If you want to read the output of this pipeline (.set file) in R using R.matlab::readMat (or other functions that internally call this function, such as eeguana::read_set), then the option 'If set, write files in Matlab v6.5...' in EEGLAB -> Preferences must be set to true!

Reference