EEG During Mental Arithmetic Tasks (EEGMAT)

Description

This is the EEGMAT (“EEG During Mental Arithmetic Tasks”) database — resting and task EEG recorded from healthy university students before and during a mental serial-subtraction task (Zyma et al., 2019). It was collected at the Educational and Scientific Centre “Institute of Biology and Medicine,” National Taras Shevchenko University of Kyiv (Ukraine) to support investigation of EEG Fourier power-spectral, coherence, and detrended-fluctuation characteristics under cognitive workload. The released cohort is 36 healthy volunteers of matched age, drawn from a larger study sample of 66 participants (47 women, 19 men); 30 were excluded from the public release for poor EEG quality (Zyma et al., 2019). For each subject the database provides a background (eyes-closed rest) recording and an arithmetic-task recording, making it a widely used benchmark for mental-workload and cognitive-load EEG classification.

Citation

• Paper DOI: 10.3390/data4010014
• Dataset DOI: 10.13026/C2JQ1P
• Reference: Zyma, I., Tukaev, S., Seleznov, I., Kiyono, K., Popov, A., Chernykh, M., & Shpenkov, O. (2019). Electroencephalograms during Mental Arithmetic Task Performance. Data 4(1), 14. Dataset: Zyma, I., Tukaev, S., & Seleznov, I. (2019). EEG During Mental Arithmetic Tasks (v1.0.0). PhysioNet.

Catalog citation block

Zyma, I., Tukaev, S., Seleznov, I., Kiyono, K., Popov, A., Chernykh, M., & Shpenkov, O. (2019). Electroencephalograms during Mental Arithmetic Task Performance. Data, 4(1), 14. https://doi.org/10.3390/data4010014

Dataset: Zyma, I., Tukaev, S., & Seleznov, I. (2019). EEG During Mental Arithmetic Tasks (version 1.0.0). PhysioNet. https://doi.org/10.13026/C2JQ1P

Companion: Zyma, I., et al. (2019). Detrended Fluctuation, Coherence, and Spectral Power Analysis of Activation Rearrangement in EEG Dynamics During Cognitive Workload. Frontiers in Human Neuroscience, 13, 270.

Download and access

Field	Value
Official source	https://physionet.org/content/eegmat/1.0.0/
Access class	open
Size	TODO(confirm)
Format	edf, csv
BIDS	false
Mirrors	none recorded

Source & access

• Repository: PhysioNet — EEG During Mental Arithmetic Tasks v1.0.0 (originally PhysioBank).
• PhysioNet DOI: 10.13026/C2JQ1P.
• Data-descriptor paper: Zyma et al. (2019), Data 4(1):14, 10.3390/data4010014 (open access, MDPI).
• License: Open Data Commons Attribution License v1.0 (ODC-By 1.0).
• Companion analysis paper: Zyma et al. (2019), Frontiers in Human Neuroscience 13:270 — detrended-fluctuation, coherence, and spectral-power analysis of the same recordings.

Data structure

The collection is distributed as EDF files, two per subject: a file with the _1 postfix = the background EEG (before the task) and a file with the _2 postfix = the EEG during mental arithmetic. A top-level subject-info.csv carries the demographic/performance metadata described above. This is not a BIDS dataset and there is no DataLad/git-annex layer in the upstream PhysioNet distribution.

License

• Name: ODC-By-1.0
• Snippets on this site: allowed — the site may ship short derived snippets and precomputed products from this dataset, labelled as derivatives (re-referenced, filtered, cropped) with citation, license and DOI.

Acquisition

Field	Value
Device	Neurocom EEG 23-channel system (XAI-MEDICA), Ag/AgCl electrodes, 19 released 10-20 channels
Device class	research-cap
Channels	19
Sampling rate (Hz)	500
Online filters	hardware 50 Hz notch; a filter the descriptor calls a “30 Hz cut-off” (read as a low-pass roll-off near 30 Hz; TODO(confirm)); ICA applied upstream
Reference	linked (interconnected) ear electrodes
Mains frequency (Hz)	50
Paradigms	rest-ec, mental-arithmetic
Participants	36 (36 official participants (PhysioNet); 41 file prefixes)
Population	36 healthy students (24 “good” and 12 “bad” counters); demographics in subject-info.csv
Clinical groups	none
Sessions	1
Durations	60-s eyes-closed rest (_1) and 60-s serial subtraction (_2) per subject

Acquisition

Recordings were made on a Neurocom EEG 23-channel system (Ukraine, XAI-MEDICA), recorded monopolarly with Ag/AgCl electrodes placed per the international 10-20 scheme and referenced to interconnected ear electrodes. The native sampling rate is 500 Hz (per the data descriptor and PhysioNet). A 50 Hz power-line notch filter was applied in hardware, alongside a filter the paper reports with a “30 Hz cut-off” — given the band of interest this is best read as a low-pass roll-off near 30 Hz rather than a 30 Hz high-pass (the wording in the descriptor is ambiguous; treat the usable passband as roughly sub-30 Hz at the top). At the preprocessing stage Independent Component Analysis (ICA) was used to remove eye, muscle, and cardiac (pulsation) artifacts, and only artifact-free 60-second segments are released.

Participants

The public release contains 36 subjects (NOT 41 — see caveats). Performance on the serial-subtraction task split them into two groups recorded in subject-info.csv:

• Group G (good counters): 24 subjects, mean 21 subtraction operations per 4 minutes (SD = 7.4).
• Group B (bad counters / “poor counters”): 12 subjects, mean 7 operations per 4 minutes (SD = 3.6), for whom the task required excessive effort.

All participants were healthy students of matched age; the accompanying subject-info.csv lists per-subject count quality (0 = Group B, 1 = Group G), gender, age, job, and recording date. The source study began with 66 participants (47 women, 19 men) and excluded 30 for EEG artifact before release (Zyma et al., 2019). No clinical group is included — this is an entirely healthy-volunteer cohort.

Tasks / conditions

The paradigm is mental serial subtraction: subjects repeatedly subtracted one number from another in their heads (e.g., subtracting a two-digit number from a four-digit number). There are exactly two conditions per subject — a background/reference EEG before the task and the EEG recorded during the arithmetic task. There is no meditation, no separate baseline/rest taxonomy, and no multi-experiment structure beyond this background-vs-task split.

Used on this site

L1.1, L1.3, L1.6

Loader

Fetches only the stated subset. Recorded loader: HTTPS (PhysioNet) — see snippet.

# PhysioNet distributes the EDF files directly over HTTPS.
# One subject: background (_1) and arithmetic (_2) recordings.
wget https://physionet.org/files/eegmat/1.0.0/Subject00_1.edf \
     https://physionet.org/files/eegmat/1.0.0/Subject00_2.edf \
     https://physionet.org/files/eegmat/1.0.0/subject-info.csv

Caveats worth knowing

• Hardware 50 Hz notch and a reported “30 Hz cut-off” (read as a low-pass; the descriptor’s wording is ambiguous).
• ICA applied upstream, so removed components are unrecoverable.
• PhysioNet N = 36, not the 41 file prefixes.

Notable caveats

• Subject count: the authoritative N is 36 (24 good + 12 bad counters). Reconcile against the per-subject roster / subject-info.csv before quoting an N.
• Channels: the device is a 23-channel system; the released EEG montage is the standard 19 10-20 electrodes (Fp1, Fp2, F7, F3, Fz, F4, F8, T7, C3, Cz, C4, T8, P7, P3, Pz, P4, P8, O1, O2). The remaining channels are reference/auxiliary, not cortical EEG.
• Line noise: the 50 Hz notch is a hardware notch hole; do not double-notch in software.
• ICA already applied: artifacts were removed upstream with ICA, so removed components are not recoverable, and each release segment is only 60 s of cleaned signal.

Related datasets

• ERP CORE (Compendium of Open Resources and Experiments)
• EEG Motor Movement/Imagery Dataset (EEGMMIDB)
• LEMON (MPI-Leipzig Mind-Brain-Body)
• Dortmund Vital Study resting EEG
• SRM Resting-state EEG

# Loader recorded in the catalog: HTTPS (PhysioNet) — see snippet
# TODO(confirm) the exact call and the subset to fetch. Official source: https://physionet.org/content/eegmat/1.0.0/