These MATLAB scripts were used for a project measuring neural entrainment using EEG. I have put them here with the hope that they will be useful to others interested in analyzing EEG data. They can be used as a template for another analysis or for inspiration. Many of the scripts can be used outside of this project (see the General EEG Analysis section).
The analysis folder is a combination of wrapper scripts for EEGLAB (mostly for preprocessing), and custom functions for selecting independent components by location and measuring entrainment. The task folder is for presenting stimuli on a Windows computer, and was used to send port codes to a BioSemi EEG recording system.
Coming soon...
The function getpath is used to access all required directory paths and files for the rest of the scripts in this toolbox (except for the General EEG Analysis functions, which don't require any external paths or files). Paths can be edited, added, or removed from that function as needed. Here is the folder structure for the current project, with some example filenames:
project_folder/ set this in getroot
├─ analysis/ the folder from this repository
│ └─ en_diary.csv this file should be edited with your own data
│ └─ en_log.csv this file should be edited with your own data
│
├─ data/
│ ├─ bdf/ .bdf files of raw EEG recordings
│ ├─ eeg_preprocessed/ .set files from en_preprocess_eeg
│ ├─ eeg_goodcomps/ .png topo- and dip-plots from en_entrainment_eeg
│ │ └─ pmc subfolder(s) for specific brain regions
│ ├─ eeg_entrainment/ .csv files from en_entrainment_eeg
│ ├─ eeg_topoplots/ .fig and .png topoplots files from en_preprocess_eeg
│ ├─ logfiles/ .csv and .txt files should be copied from task/logfiles
│ ├─ midi/ .mid and .wav tapping files exported from Pro Tools
│ └─ tapping/ .mat files from en_preprocess_tapping
│
├─ eeglab/ EEGLAB toolbox (https://bitbucket.org/sccn_eeglab/eeglab.git)
│
├─ miditoolbox1.1/ MIDI toolbox (https://github.com/miditoolbox/1.1)
│
└─ task/ from this repository; move to stimulus presentation computer
└─ logfiles/ these are just logfiles from testing
Note: the task scripts save the logfiles in the task/logfiles folder; these files should be moved to the logfiles folder for analysis.
The diary.csv file can be considered a sort of configuration file for the analysis. It should contain the following columns, with each row representing a single participant:
| Variable | Type | Description |
|---|---|---|
| id | [number] | Each row should be a unique number. When calling the functions in this toolbox, use this number to refer to a specific participant. |
| incl | [boolean; 1 or 0] | Whether or not to include this participant in the analysis. When calling the "getdata" or looping functions in this toolbox, leaving the id argument empty will use all id's that have a 1 here. |
| order | [number; 1 or 2] | This is used by the en_epoch function. |
| bdffile | [comma-separated list of strings] | List of bdf (raw BioSemi EEG data) files associated with this id. Don't include the ".bdf". If there are multiple files, en_readbdf will read them all and merge them using EEGLAB's pop_mergeset. |
| hc | [numeric] | Circumference of the participant's head in centimeters (cm). This can be used for dipole fitting by en_dipfit. |
| capsize | [string, 'M' or 'ML'] | Which BioSemi cap was used, either medium (M) or medium-large (ML). This is used by the en_dipfit function for the grid search. |
| eventchans | [number] | The channel from which the portcode information should be extracted. Used by the en_readbdf function. |
| badchans | [comma-separated list of strings] | Channel/electrode labels that should be removed because they were noisy during the recording. |
| extra_eeg_event | [comma-separated list of numbers] | If there were extra portcodes sent (e.g., because a trial was repeated and the portcode was erroneously sent twice), enter the indices of the portcodes that should be removed. For example, if there are 60 trials, and the portcode for the 30th trial was sent twice (i.e., the 30th and 31st portcodes are the same trial, and you would like to remove the first one), enter 30 here. |
| missed_eeg_event | [comma-separated list of numbers] | If some portcodes did not get sent (e.g., because the BioSemi battery died, but the experiment continued), enter the indices of the portcodes that were missed. For example, if there are 60 trials, and the battery died after the 10th trial, you didn't notice and stop the experiment until after the 15th trial (trials 11 to 15 did not have their portcodes recorded), enter "11, 12, 13, 14, 15" here (without the quotes). Note that if the battery dies in the middle of a trial, you might want to consider adding that trial to extra_eeg_event. Filling in this field will probably require some comparison of the logfile (which portcodes were sent) and the BDF file (which portcodes were recorded). |
| extra_midi_event | [comma-separated list of numbers] | Same as extra_eeg_event, but for the MIDI recording. |
| missed_midi_event | [comma-separated list of numbers] | Same as missed_eeg_event, but for the MIDI recording. |
| experimenters | [comma-separated list of strings] | Initials of the experimenters for this participant's session. |
| recording_notes | [semicolon-separated list of sentences] | Any notes from the EEG recording session that might be good to know. |
| midi_audio_marker_threshold | [number] | The threshold parameter used by findAudioMarkers when finding event markers for epoching MIDI data. If the audio level is the same for all participants than this will not be needed. Leaving this empty will use the default value set in en_preprocess_tapping. |
| midi_timeBetween_secs | [number] | The timeBetween parameter used by findAudioMarkers when finding event markers for epoching MIDI data. Make this shorter if the participant was really quick advancing from one trial to the next. leaving this empty will use the default value set in en_preprocess_tapping. |
| dipolar_comps | [comma-separated list of numbers] | After running en_preprocess_eeg, look at the topographical plots that are saved and mark down which components are dipolar. This field is used by en_entrainment_eeg to select good components with select_comps. See Delorme, Palmer, Onton, Oostenveld, & Makeig (2012; PLOS ONE) for more information. |
The following code will run the entire analysis pipeline from raw data files to a tabular data frame of entrainment values.
% setup directory structure and fill it with data files
% make a copy of en_diary_example.csv and fill it with your id numbers, filenames, etc.
% make a copy of getroot_example.m and rename it getroot.m
% edit getroot.m to return the absolute path to the project root (i.e., the "project_folder" in the directory structure above)
% set current folder to this analysis folder, or add it to the MATLAB path
en_preprocess % preprocess all EEG and tapping data for all IDs, save topoplots
% [optional] look at topographical maps saved in getpath('topoplots') and mark down component numbers that are dipolar in en_diary.csv
en_entrainment % find good, localized components and calculate entrainment
en_writetable('entrainment.csv') % write all data in tabular format to a csv fileEach function list is loosely in the order that they would be used in the processing pipeline.
Perform a whole section of the analysis pipeline and batch processing. These mostly contain calls to the other functions in the lists below.
| Function | Description |
|---|---|
en_preprocess |
Loops through the specified IDs and runs en_preprocess_eeg. All output from the command window is captured for each ID and saved as a .log file to getpath('eeg'), as well as a summary file (en_log.csv) to keep track of what has been completed and if errors were present. |
en_entrainment |
Loops through specified IDs and runs eeg_entrainment. |
en_tapping |
Loops through specified IDs and runs en_preprocess_tapping and en_entrainment_tapping. |
en_preprocess_eeg |
A macro that runs en_readbdf and a number of EEGLAB functions to pre-process EEG data, including ICA and dipole fitting. Resultant .set files are saved to getpath('eeg'). |
en_preprocess_tapping |
A macro that loads the .wav file of stimuli that was exported from Pro Tools, searches for audio onsets to get the stimulus onset times, and uses these to epoch the MIDI tapping data. MIDI data is read using MIDI Toolbox and saved as a MATLAB table. |
en_entrainment_tapping |
A macro that takes an ID, filename, or table and outputs a table of entrainment values for each trial. Resultant tables are saved as .csv files in getpath('eeg_tapping'). |
Simplify finding and loading files.
| Function | Description |
|---|---|
getroot |
This function should return the absolute path to the project root folder. It is used by getpath.m to find files and folders needed for the analysis |
getpath |
Takes a keyword as input and returns a directory or file path. All functions in this "toolbox" use this function to get path names. This means that you can move these scripts to a different computer or port them to a different project, and will only have to change this file in order for everything to work (theoretically). This function depends on getroot.m Note that you will have to create this file from the example that is provided. |
en_load |
Takes a keyword (and optionally an ID number), and loads the specified file into the MATLAB workspace. Can also start EEGLAB from the path specified in getpath('eeglab'). |
These include wrappers on EEGLAB functions (mostly EEG preprocessing) and custom scripts for spectral analyses.
| Function | Description |
|---|---|
en_readbdf |
Takes an ID number as input, gets their .bdf filenames from en_diary.csv, and reads those files into an EEGLAB .set file (if there are multiple .bdf files, they are merged). It also converts the channel labels from the BioSemi alphabetical labels into Oostenveld's 10-5 system. |
en_epoch |
Epochs an EEG struct using EEGLAB based on portcodes which are specified with keywords. This function in particular is highly specialized for the current study, and will require lots of editing to work for a different study. |
en_dipfit |
Wrapper on EEGLAB functions for dipole fitting using the Boundary Element Model (BEM). |
Some mostly EEG-related functions for preprocessing and spectral analysis that will work outside of this project.
| Function | Description |
|---|---|
alpha2fivepct |
Takes an EEG struct or a cell of strings, and remaps channel labels from BioSemi alphabetical labels to Oostenveld's 10-5 system. |
averageReference |
Performs average referencing for fully-ranked data on an EEG struct. |
tal2region |
This is a wrapper for the Talairach client. It requires that the talairach.jar file be in the same folder. |
region2comps |
This is a wrapper for tal2region, that takes an EEG struct and a brain region as input, and returns the IC numbers that are in, or close to that region. |
select_comps |
Takes a preprocessed EEG struct and returns component numbers that match three criteria: region, residual variance, or by manually giving indices. In this project, the indices input is used for manually selecting components that are dipolar. |
dtplot |
Takes a preprocessed EEG struct and displays both a topographical plot and a dipole plot for the specified components. If a folder argument is given, these plots are saved as .png files instead of being displayed. |
getfft3 |
A fancy wrapper on MATLAB's fft function that expects data to be channels-by-time-by-trials (e.g., EEGLAB's EEG.data). |
noisefloor3 |
For each value in the data, remove the mean of surrounding values. This version expects data to be channels-by-time-by-trials (e.g., EEGLAB's EEG.data). |
getbins3 |
Given some data and a vector of labels, get the value (or mean/max/min) of data for specified labels. This is used to get the max spectral value at a given frequency. |
findAudioMarkers |
Look through an audio file and return the onset times in samples. This file is adapted from PHZLAB. |
eeg_entrainment |
Takes an EEG struct and a beat frequency, and returns entrainment values for components that are localized to specified regions. |
Please see the documentation included in each of the functions in the task folder for an explanation of how they work.
These scripts were written using MATLAB R2017a, EEGLAB 14.1.1, and MIDI Toolbox 1.1.