eegdash 0.3.9.dev182388821

EEG data for machine learning

EEG-Dash

To leverage recent and ongoing advancements in large-scale computational methods and to ensure the preservation of scientific data generated from publicly funded research, the EEG-DaSh data archive will create a data-sharing resource for MEEG (EEG, MEG) data contributed by collaborators for machine learning (ML) and deep learning (DL) applications.

Data source

The data in EEG-DaSh originates from a collaboration involving 25 laboratories, encompassing 27,053 participants. This extensive collection includes MEEG data, which is a combination of EEG and MEG signals. The data is sourced from various studies conducted by these labs, involving both healthy subjects and clinical populations with conditions such as ADHD, depression, schizophrenia, dementia, autism, and psychosis. Additionally, data spans different mental states like sleep, meditation, and cognitive tasks. In addition, EEG-DaSh will incorporate a subset of the data converted from NEMAR, which includes 330 MEEG BIDS-formatted datasets, further expanding the archive with well-curated, standardized neuroelectromagnetic data.

Featured data

The following HBN datasets are currently featured on EEGDash. Documentation about these datasets is available here.

DatasetID	Participants	Files	Sessions	Population	Channels	Is 10-20?	Modality	Size
ds005505	136	5393	1	Healthy	129	other	Visual	103 GB
ds005506	150	5645	1	Healthy	129	other	Visual	112 GB
ds005507	184	7273	1	Healthy	129	other	Visual	140 GB
ds005508	324	13393	1	Healthy	129	other	Visual	230 GB
ds005510	135	4933	1	Healthy	129	other	Visual	91 GB
ds005512	257	9305	1	Healthy	129	other	Visual	157 GB
ds005514	295	11565	1	Healthy	129	other	Visual	185 GB

A total of 246 other datasets are also available through EEGDash.

Data format

EEGDash queries return a Pytorch Dataset formatted to facilitate machine learning (ML) and deep learning (DL) applications. PyTorch Datasets are the best format for EEGDash queries because they provide an efficient, scalable, and flexible structure for machine learning (ML) and deep learning (DL) applications. They allow seamless integration with PyTorch’s DataLoader, enabling efficient batching, shuffling, and parallel data loading, which is essential for training deep learning models on large EEG datasets.

Data preprocessing

EEGDash datasets are processed using the popular braindecode library. In fact, EEGDash datasets are braindecode datasets, which are themselves PyTorch datasets. This means that any preprocessing possible on braindecode datasets is also possible on EEGDash datasets. Refer to braindecode tutorials for guidance on preprocessing EEG data.

EEG-Dash usage

Install

Use your preferred Python environment manager with Python > 3.9 to install the package.

• To install the eegdash package, use the following command: pip install eegdash
• To verify the installation, start a Python session and type: from eegdash import EEGDash

Data access

To use the data from a single subject, enter:

from eegdash import EEGDashDataset

ds_NDARDB033FW5 = EEGDashDataset(
    {"dataset": "ds005514", "task":
     "RestingState", "subject": "NDARDB033FW5"}, 
     cache_dir="."
)

This will search and download the metadata for the task RestingState for subject NDARDB033FW5 in BIDS dataset ds005514. The actual data will not be downloaded at this stage. Following standard practice, data is only downloaded once it is processed. The ds_NDARDB033FW5 object is a fully functional braindecode dataset, which is itself a PyTorch dataset. This tutorial shows how to preprocess the EEG data, extracting portions of the data containing eyes-open and eyes-closed segments, then perform eyes-open vs. eyes-closed classification using a (shallow) deep-learning model.

To use the data from multiple subjects, enter:

from eegdash import EEGDashDataset

ds_ds005505rest = EEGDashDataset(
    {"dataset": "ds005505", "task": "RestingState"}, target_name="sex", cache_dir=".
)

This will search and download the metadata for the task 'RestingState' for all subjects in BIDS dataset 'ds005505' (a total of 136). As above, the actual data will not be downloaded at this stage so this command is quick to execute. Also, the target class for each subject is assigned using the target_name parameter. This means that this object is ready to be directly fed to a deep learning model, although the tutorial script performs minimal processing on it, prior to training a deep-learning model. Because 14 gigabytes of data are downloaded, this tutorial takes about 10 minutes to execute.

Automatic caching

By default, EEGDash caches downloaded data under a single, consistent folder:

• If EEGDASH_CACHE_DIR is set in your environment, that path is used.
• Else, if MNE’s MNE_DATA config is set, that path is used to align with other EEG tooling.
• Otherwise, .eegdash_cache in the current working directory is used.

This means that if you run the tutorial scripts, the data will only be downloaded the first time the script is executed and reused thereafter.

Education -- Coming soon...

We organize workshops and educational events to foster cross-cultural education and student training, offering both online and in-person opportunities in collaboration with US and Israeli partners. Events for 2025 will be announced via the EEGLABNEWS mailing list. Be sure to subscribe.

About EEG-DaSh

EEG-DaSh is a collaborative initiative between the United States and Israel, supported by the National Science Foundation (NSF). The partnership brings together experts from the Swartz Center for Computational Neuroscience (SCCN) at the University of California San Diego (UCSD) and Ben-Gurion University (BGU) in Israel.