SSVEP MAMEM 2 dataset

SSVEP MAMEM 2 dataset.

Dataset Overview

• Code: MAMEM2
• Paradigm: ssvep
• DOI: 10.48550/arXiv.1602.00904
• Subjects: 11
• Sessions per subject: 1
• Events: 6.66=1, 7.50=2, 8.57=3, 10.00=4, 12.00=5
• Trial interval: [1, 4] s
• Runs per session: 5
• File format: MAT

Acquisition

• Sampling rate: 250.0 Hz
• Number of channels: 256
• Channel types: eeg=256
• Channel names: E1, E10, E100, E101, E102, E103, E104, E105, E106, E107, E108, E109, E11, E110, E111, E112, E113, E114, E115, E116, E117, E118, E119, E12, E120, E121, E122, E123, E124, E125, E126, E127, E128, E129, E13, E130, E131, E132, E133, E134, E135, E136, E137, E138, E139, E14, E140, E141, E142, E143, E144, E145, E146, E147, E148, E149, E15, E150, E151, E152, E153, E154, E155, E156, E157, E158, E159, E16, E160, E161, E162, E163, E164, E165, E166, E167, E168, E169, E17, E170, E171, E172, E173, E174, E175, E176, E177, E178, E179, E18, E180, E181, E182, E183, E184, E185, E186, E187, E188, E189, E19, E190, E191, E192, E193, E194, E195, E196, E197, E198, E199, E2, E20, E200, E201, E202, E203, E204, E205, E206, E207, E208, E209, E21, E210, E211, E212, E213, E214, E215, E216, E217, E218, E219, E22, E220, E221, E222, E223, E224, E225, E226, E227, E228, E229, E23, E230, E231, E232, E233, E234, E235, E236, E237, E238, E239, E24, E240, E241, E242, E243, E244, E245, E246, E247, E248, E249, E25, E250, E251, E252, E253, E254, E255, E256, E26, E27, E28, E29, E3, E30, E31, E32, E33, E34, E35, E36, E37, E38, E39, E4, E40, E41, E42, E43, E44, E45, E46, E47, E48, E49, E5, E50, E51, E52, E53, E54, E55, E56, E57, E58, E59, E6, E60, E61, E62, E63, E64, E65, E66, E67, E68, E69, E7, E70, E71, E72, E73, E74, E75, E76, E77, E78, E79, E8, E80, E81, E82, E83, E84, E85, E86, E87, E88, E89, E9, E90, E91, E92, E93, E94, E95, E96, E97, E98, E99
• Montage: GSN-HydroCel-256
• Hardware: EGI 300 Geodesic EEG System (GES 300)
• Reference: Cz
• Line frequency: 50.0 Hz
• Impedance threshold: 80.0 kOhm
• Cap manufacturer: EGI
• Cap model: HydroCel Geodesic Sensor Net (HCGSN)

Participants

• Number of subjects: 11
• Health status: healthy
• Age: min=24, max=39
• Gender distribution: male=8, female=3
• Handedness: {'right': 10, 'left': 1}

Experimental Protocol

• Paradigm: ssvep
• Number of classes: 5
• Class labels: 6.66, 7.50, 8.57, 10.00, 12.00
• Trial duration: 5.0 s
• Study design: Subjects focus attention on visual stimuli flickering at different frequencies (6.66, 7.50, 8.57, 10.00, 12.00 Hz) to select commands. Each stimulus presented for 5 seconds followed by 5 seconds rest.
• Feedback type: none
• Stimulus type: flickering box
• Stimulus modalities: visual
• Primary modality: visual
• Synchronicity: synchronous
• Mode: offline
• Stimulus presentation: SoftwareName=Microsoft Visual Studio 2010 with OpenGL, device=22 inch LCD monitor, refresh_rate=60 Hz, resolution=1680x1080

HED Event Annotations

Schema: HED 8.4.0 | Browse: https://www.hedtags.org/hed-schema-browser

  6.66
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/6_66

  7.50
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/7_50

  8.57
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/8_57

  10.00
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/10_00

  12.00
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/12_00

Paradigm-Specific Parameters

• Detected paradigm: ssvep
• Stimulus frequencies: [6.66, 7.5, 8.57, 10.0, 12.0] Hz
• Number of targets: 5
• Number of repetitions: 3

Data Structure

• Trials: 1104
• Trials context: Each session includes 23 trials (8 adaptation trials excluded from analysis). 5 sessions per subject (with exceptions: S001=3 sessions, S003=4 sessions, S004=4 sessions). Total: 1104 trials of 5 seconds each.

Preprocessing

• Data state: raw
• Preprocessing applied: False

Signal Processing

• Classifiers: LDA, SVM, Random Forest, kNN, Naive Bayes, AdaBoost, Decision Trees, CCA
• Feature extraction: PWelch, Periodogram, FFT, Goertzel, PYULEAR (Yule-AR), STFT, DWT, PSD, Wavelet, Spectrogram
• Frequency bands: analyzed=[5.0, 48.0] Hz
• Spatial filters: CAR, CSP, Minimum Energy

Cross-Validation

• Method: leave-one-subject-out
• Evaluation type: cross_subject

Performance (Original Study)

• Accuracy: 74.42%
• Mean Accuracy Default Config: 72.47
• Mean Accuracy Optimal Config: 74.42
• Processing Time Msec: 68

BCI Application

• Applications: command_selection
• Environment: laboratory
• Online feedback: False

Tags

• Pathology: Healthy
• Modality: Visual
• Type: Research

Documentation

• DOI: 10.48550/arXiv.1602.00904
• Associated paper DOI: arXiv:1602.00904v2
• License: ODC-By-1.0
• Investigators: Vangelis P. Oikonomou, Georgios Liaros, Kostantinos Georgiadis, Elisavet Chatzilari, Katerina Adam, Spiros Nikolopoulos, Ioannis Kompatsiaris
• Institution: Centre for Research and Technology Hellas (CERTH)
• Country: GR
• Repository: GitHub
• Data URL: https://figshare.com/articles/dataset/3153409
• Publication year: 2016
• Funding: H2020-ICT-2014-644780
• Ethics approval: Approved by ethics committee of Centre for Research and Technology Hellas, date 3/7/2015, grant H2020-ICT-2014-644780
• Keywords: SSVEP, BCI, brain-computer interface, EEG, visual evoked potentials, signal processing, feature extraction, classification

Abstract

Brain-computer interfaces (BCIs) have been gaining momentum in making human-computer interaction more natural, especially for people with neuro-muscular disabilities. This study focuses on SSVEP-based BCIs and performs a comparative evaluation of state-of-the-art algorithms for filtering, artifact removal, feature extraction, feature selection and classification. Dataset consists of 256-channel EEG signals from 11 subjects with 5 flickering frequencies (6.66, 7.50, 8.57, 10.00, 12.00 Hz).

Methodology

Leave-one-subject-out cross-validation was used to evaluate a general-purpose BCI system without subject-specific training. Systematic comparison of algorithms across all signal processing stages: (1) Signal filtering: FIR vs IIR filters; (2) Artifact removal: AMUSE vs FastICA; (3) Feature extraction: PWelch, Periodogram, PYULEAR, DWT, STFT, Goertzel; (4) Feature selection: entropy-based methods and PCA/SVD; (5) Classification: SVM, LDA, KNN, Naive Bayes, Random Forest, AdaBoost. Optimal configuration achieved 74.42% mean accuracy using IIR-Elliptic filter, AMUSE artifact removal, PWelch feature extraction with nfft=512, segment length=350, overlap=0.75, and channel-138.

References

Oikonomou, V. P., Liaros, G., Georgiadis, K., Chatzilari, E., Adam, K., Nikolopoulos, S., & Kompatsiaris, I. (2016). Comparative evaluation of state-of-the-art algorithms for SSVEP-based BCIs. arXiv preprint arXiv:1602.00904.

MAMEM Steady State Visually Evoked Potential EEG Database <https://archive.physionet.org/physiobank/database/mssvepdb/>_

S. Nikolopoulos, 2016, DataAcquisitionDetails.pdf <https://figshare.com/articles/dataset/MAMEMEEGSSVEPDatasetII256channels11subjects5frequenciespresentedsimultaneously/3153409?file=4911931> Appelhoff, S., Sanderson, M., Brooks, T., Vliet, M., Quentin, R., Holdgraf, C., Chaumon, M., Mikulan, E., Tavabi, K., Hochenberger, R., Welke, D., Brunner, C., Rockhill, A., Larson, E., Gramfort, A. and Jas, M. (2019). MNE-BIDS: Organizing electrophysiological data into the BIDS format and facilitating their analysis. Journal of Open Source Software 4: (1896). https://doi.org/10.21105/joss.01896

Pernet, C. R., Appelhoff, S., Gorgolewski, K. J., Flandin, G., Phillips, C., Delorme, A., Oostenveld, R. (2019). EEG-BIDS, an extension to the brain imaging data structure for electroencephalography. Scientific Data, 6, 103. https://doi.org/10.1038/s41597-019-0104-8

---

Generated by MOABB 1.4.3 (Mother of All BCI Benchmarks) https://github.com/NeuroTechX/moabb