Using GEDAI for online (real-time) denoising

This tutorial demonstrates how to use GEDAI for online (real-time) denoising.

from mne import make_fixed_length_epochs
from mne.io import read_raw

from gedai import Gedai
from gedai.data import get_contaminated_eeg_set_path

raw = read_raw(str(get_contaminated_eeg_set_path()), preload=True)

# Crop to the first 15 seconds for demonstration purposes
# (Remove or adjust this for full data analysis)
raw.crop(0, 30)

# Apply average reference (standard preprocessing for EEG)
raw.set_eeg_reference("average", projection=False)

EEG channel type selected for re-referencing
Applying average reference.
Applying a custom ('EEG',) reference.

	General
	Filename(s)	SNR=0.35481 contamination=25 clean_EEG_dataset_2.set + EOG_EMG_NOISE_dataset_1.set
	MNE object type	RawEEGLAB
	Measurement date	Unknown
	Participant	Unknown
	Experimenter	Unknown
	Acquisition
	Duration	00:00:31 (HH:MM:SS)
	Sampling frequency	200.00 Hz
	Time points	6,001
	Channels
	EEG	27
	Head & sensor digitization	30 points
	Filters
	Highpass	0.00 Hz
	Lowpass	100.00 Hz

For this example, we will use the standard broadband GEDAI. However, the same principles apply to the spectral GEDAI as well. For real-time applications, data is typically processed in chunks (e.g., sliding windows). To simulate this, we will first convert the raw data into overlapping epochs.

epochs = make_fixed_length_epochs(raw, duration=1.0, overlap=0.5, preload=True)

Not setting metadata
59 matching events found
No baseline correction applied
0 projection items activated
Using data from preloaded Raw for 59 events and 200 original time points ...
0 bad epochs dropped

We first need to fit the GEDAI model on an initial data segment. This segment should be representative of the data to be denoised, including typical artifacts. A common approach is to use a baseline period at the beginning of the recording for this purpose.

gedai = Gedai()

# In this example, we use the first 10 seconds (i.e., first 20 epochs) for
# model fitting. The baseline period
baseline_epochs = epochs[:20]
gedai.fit_epochs(baseline_epochs, verbose=True)

343 x 343 full covariance (kind = 1) found.

After fitting, we can apply the fitted GEDAI model to each incoming epoch for denoising.

for epoch_idx in range(20, len(epochs)):
    epoch = epochs[epoch_idx : epoch_idx + 1]  # Get the current epoch
    denoised_epoch = gedai.transform_epochs(epoch)

Estimated memory usage: 299 MB