39. Sample data: EEG‐MRI during music listening

This is the concurrent EEG-MRI data for the music listening experiment, which has been performed on epilepsy patients with invasive recording. Details about the background and the stimulus can be referred here. Note that participants for this data set are different from epilepsy patients in the invasive recording.

Data

MRI data were acquired on a 3T MRI systems (Skyra, Siemens) with a 32-channel whole-head coil array. Structural images were acquired with MPRAGE in a 1-mm isotropic resolution before and after electrode implantation. EEG data acquired using a 32-electrode cap (Brain Products). The followings are links to files to eight subjects (s001 - s008).

EEG data	fMRI data	MRI (FreeSurfer)
tar ball	tar ball	tar ball

Experiment design

Here is the table listing the song played in each run. "D1" means the first play of song "D". There songs ("D", "B", and "L") were played twice.

participant	D1	B1	L1	D2	B2	L2	Resting
s001	3	2	4	6	5	7	1
s002	2	4	3	5	7	6	1
s003	4	3	2	7	6	5	1
s004	3	4	2	6	7	5	1
s005	2	3	4	5	6	7	1
s006	4	2	3	7	5	6	1
s007	3	2	4	6	5	7	1
s008	2	4	3	5	7	6	1

Analysis

EEG data

EEG data are stored in triplets of .vmrk, .vhdr, and .eeg files. Use bvloader to read data.

Example codes:

headerFile = {
    '../eeg_raw/eyeclose_NOMR.vhdr';
    '../eeg_raw/eyeopen_NOMR.vhdr';
    };

markerFile={
    '../eeg_raw/eyeclose_NOMR.vmrk';
    '../eeg_raw/eyeopen_NOMR.vmrk';
    };

for f_idx=1:length(headerFile)
% first get the continuous data as a matlab array
    eeg{f_idx} = double(bva_loadeeg(headerFile{f_idx}));
    
    % meta information such as samplingRate (fs), labels, etc
    [fs(f_idx) label meta] = bva_readheader(headerFile{f_idx});

    % markers..
    trigger{f_idx}=etc_read_vmrk(markerFile{f_idx});
end;

    

EEG data collected inside MRI are contaminated by MRI gradient coil switching and participant's heartbeats. Refer to this page for suppression of these ballistocardiography (BCG) artifacts.

fMRI data

read STC files

STC files are two-dimensional data array to store signals across different positions (rows) at different time instants (columns).

Please refer to this page by Matlab and this page by Python.

All fMRI data files have been trimmed so that their duration are identical across participants for the same played musical pieces.

Tasks

Task 1: Inter-subject correlation maps for fMRI signals.

Calculate the synchronized fMRI signals across participants using pre-processed data in a common atlas space (MNI305). Here explains the inter-subject correlation (ISC) analysis in fMRI.

• Task 1.1. Calculate the inter-subject correlation maps for each song
• Task 1.2. Use linear model to estimate the ISC effects of "song listening" and "difference in repeated listening".

Task 2: Suppression of EEG artifacts

Examine and suppress MRI gradient and pulse artifacts. Here is the link about BCG artifacts suppression.

• Task 2.1. Reduce the gradient artifacts and examine how EEG waveforms look alike.
• Task 2.2. Reduce the pulse artifacts and examine how EEG waveforms look alike.

Task 3: Inter-subject correlation topologies for EEG signals

Calculate the inter-subject correlation of spectral envelopes at each EEG electrode at different frequencies (4 Hz - 100 Hz). Use Morlet wavelet to calculate the envelope of oscillatory signals over time at different frequencies.

• Task 3.1. Familiarize with the Morlet wavelet transformation.
• Task 3.2. Calculate the difference of oscillatory powers between song listening.

Task 4: Inter-subject correlation maps for EEG source signals

Source modeling of each participant's EEG data. Then calculate the inter-subject correlation of spectral envelopes. Here is the link about EEG source analysis.

Task 5. Musical features vs. brain activity

Extract musical features using, for example, MIR toolbox. Correlate these features with fMRI and EEG signals.