info/tutorials/modality/processing-eeg-with-mne.md

Processing EEG with MNE-Python

import mne
import warnings
import numpy as np
import matplotlib.pyplot as plt

from almirah import Layout

mne.set_log_level(False)
warnings.filterwarnings('ignore')

lay = Layout(root="/path/to/data", specification_name="bids")
lay

<Layout root: '/path/to/data'>

lay = Layout.get(specification_name='bids')

files = lay.query(datatype="eeg", extension=".vhdr")
len(files)

2223

vhdr_file = lay.query(subject="D0019", session="101", datatype="eeg", task="rest", extension =".vhdr")[0]
eeg_file = lay.query(subject="D0019", session="101", datatype="eeg", task="rest", extension=".eeg")[0]
montage_file = lay.query(subject="D0019", session="101", space="CapTrak", suffix="electrodes")[0]

print(vhdr_file.rel_path)

sub-D0019/ses-101/eeg/sub-D0019_ses-101_task-rest_run-01_eeg.vhdr

eeg_file.download()

montage = mne.channels.read_custom_montage(montage_file.path)
raw = mne.io.read_raw_brainvision(vhdr_file.path, preload=True)
raw.set_montage(montage)
raw.info["bads"] = ["VREF"]

raw.info

<summary><strong>General</strong></summary>
<table class="table table-hover table-striped table-sm table-responsive small">
    <tr>
        <th>Measurement date</th>
        <td>Unknown</td>
    </tr>
    <tr>
        <th>Experimenter</th>
        <td>Unknown</td>
    </tr>
    <tr>
        <th>Participant</th>
        <td>Unknown</td>
    </tr>
</table>
</details>
<details open>
    <summary><strong>Channels</strong></summary>
    <table class="table table-hover table-striped table-sm table-responsive small">
        <tr>
            <th>Digitized points</th>
            <td>132 points</td>
        </tr>
        <tr>
            <th>Good channels</th>
            <td>128 EEG, 5 misc</td>
        </tr>
        <tr>
            <th>Bad channels</th>
            <td>VREF</td>
        </tr>
        <tr>
            <th>EOG channels</th>
            <td>Not available</td>
        </tr>
        <tr>
            <th>ECG channels</th>
            <td>Not available</td>
        </tr>
    </table>
    </details>
    <details open>
        <summary><strong>Data</strong></summary>
        <table class="table table-hover table-striped table-sm table-responsive small">

            <tr>
                <th>Sampling frequency</th>
                <td>1000.00 Hz</td>
            </tr>
            <tr>
                <th>Highpass</th>
                <td>0.00 Hz</td>
            </tr>
            <tr>
                <th>Lowpass</th>
                <td>500.00 Hz</td>
            </tr>
        </table>
        </details>

Preprocessing

# Apply a band-pass filter
raw.filter(1, 40, fir_design="firwin")

# Plot the raw data
raw.plot(n_channels=30, duration=30, scalings="auto")
plt.close()

Artifact removal using ICA

# Setup ICA
ica = mne.preprocessing.ICA(n_components=15, random_state=97)
ica.fit(raw)

# Plot ICA components
ica.plot_components()

# Select bad components manually (usually by visual inspection)
ica.exclude = [0, 6, 14]  # Example: components 0 and 1 are eye blinks and heartbeats

# Apply ICA removal
raw_clean = ica.apply(raw.copy())

Power Spectral Analysis

# Compute and plot Power Spectral Density (PSD)
raw_clean.compute_psd().plot(exclude="bads", amplitude=False)
plt.show()

raw_clean.compute_psd().plot_topomap(ch_type="eeg", agg_fun=np.median)
plt.close()