This is a series of scripts developed by myself for the Koen Lab to process EEG data. These are Python scripts that use the MNE-Python as well as custom function.
At present, these are primarily for pre-processing of single subject data. No group analysis scripts are included in this template.
The code is accompanied by example data store with Git LFS.
To use this template, fork this repository and create a new repository using this as a template. You will then go through the scripts and update information where needed.
The environment used for these scripts can be installed using anaconda by importing the mne_environment.yml file.
The data are imported from how they are stored and into the Brain Imaging Data Structure (BIDS) using mne-bids.
The main preprocessing options are locating in the config.py file. These options can be modified there and control all other aspects of processing. These control aspects of processing such as:
- Epoch length
- Filtering methods
- Automated bad channel and epoch methods
- ERP plotting options for reports
- Directories
This script handles importing the data (both EEG and behavioral data) from a source data folder into the BIDS format. This script is not standardized and must be updated with proper file names specific to your experiment. Relevant variables for this script located in the config.py are:
cols_to_keepcols_to_renamecols_to_add
This script will need to be customized for each experiment. The purpose is to derive the behavioral dependent variables of interest from the behavioral tasks.
This script filters the continuous data, segments the data into epochs, and performs semi-automated artifact rejection prior to running ICA estimation.
This script loads the estimated ICA components output from Step 1 and plots the component properties (using mne.preprocessing.ICA.plot_components) to be inspected for rejection. Some components are automatically flagged based on algorithms implemented in [me-faster](https://github.com/wmvanvliet/mne-faster. The components automatically flagged are not necessarily artifcatual, and require confirmation via visual inspection to mark for rejection.
This script first loads the EEG data, filters the continuous data, and segments the data into epochs. The ICA components flagged as artifactual are subtracted from the newly created epoched data and then channels marked as bad are interpolated. Then, a semi-automated artifct detection routine is implmented that does the following (in order):
- Reject epochs with a blink at stimulus onset
- Reject epochs flagged as artifacts based on mne-faster algorithms
- Visual inspection of the remaining artifacts.
During the final visual inspection, additional EEG channels can be marked as artifactual, which are subsequently interpolated. The cleaned epochs are written to file.
This script makes the ERPs and plots some for inspection of data quality. The ERPs that are written and shown can be controlled by changing the erp_queries, erp_contrasts, and erps_to_plot variables in config.py.
This script makes an html report using the Report class from mne-python. This follows a standardized format that will largely read from the above scripts output. It will need some modification for naming of the behavioral data files that are written (including figures).
The functions script contains utility functions used throughout the scripts, such as get_sub_list and adjust_events_photosensor.
This codebase is under development and may experience constant change. For this reason, it is import to use this as a template only when starting a new experiment. Do not update the 'version' of this code base for an ungoing experiment unless otherwise necessary
At present, the scripts do not contain a standard method for time-frequency analysis of EEG data. This will be developed in a future iteration of the template.