c-VEP and Burst-VEP dataset from Castillos et al. (2023)

c-VEP and Burst-VEP dataset from Castillos et al. (2023)

Dataset Overview

• Code: CastillosBurstVEP100
• Paradigm: cvep
• DOI: https://doi.org/10.1016/j.neuroimage.2023.120446
• Subjects: 12
• Sessions per subject: 1
• Events: 0=100, 1=101
• Trial interval: (0, 0.25) s
• File format: EEGLAB .set
• Number of contributing labs: 1

Acquisition

• Sampling rate: 500.0 Hz
• Number of channels: 32
• Channel types: eeg=32
• Channel names: C3, C4, CP1, CP2, CP5, CP6, Cz, F10, F3, F4, F7, F8, F9, FC1, FC2, FC5, FC6, Fp1, Fp2, Fz, O1, O2, Oz, P10, P3, P4, P7, P8, P9, Pz, T7, T8
• Montage: standard_1020
• Hardware: BrainProducts LiveAmp 32
• Reference: FCz
• Ground: FPz
• Sensor type: eeg
• Line frequency: 50.0 Hz
• Online filters: {'notch': {'freq': 50.0, 'bandwidth': 0.2, 'order': 16, 'type': 'IIR cut-band'}}
• Impedance threshold: 25.0 kOhm
• Cap manufacturer: BrainProducts
• Cap model: Acticap
• Electrode type: active

Participants

• Number of subjects: 12
• Health status: healthy
• Age: mean=30.6, std=7.1
• Gender distribution: female=4, male=8
• Species: human

Experimental Protocol

• Paradigm: cvep
• Task type: target selection
• Number of classes: 2
• Class labels: 0, 1
• Trial duration: 2.2 s
• Tasks: visual attention, target selection
• Study design: factorial within-subject
• Study domain: BCI performance and user experience
• Feedback type: none
• Stimulus type: visual
• Stimulus modalities: visual
• Primary modality: visual
• Synchronicity: synchronous
• Mode: offline
• Training/test split: False
• Instructions: Focus on cued targets sequentially in random order
• Stimulus presentation: software=PsychoPy, monitor=Dell P2419HC, resolution=1920x1080, refresh_rate_hz=60

HED Event Annotations

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

  0
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/intensity_0

  1
    ├─ Sensory-event
    ├─ Experimental-stimulus
    ├─ Visual-presentation
    └─ Label/intensity_1

Paradigm-Specific Parameters

• Detected paradigm: cvep
• Code type: burst
• Number of targets: 4
• Cue duration: 0.5 s

Data Structure

• Trials: 60
• Blocks per session: 15
• Trials context: 15 blocks x 4 trials per block = 60 trials per subject for burst c-VEP at 100% amplitude

Preprocessing

• Data state: raw

Signal Processing

• Classifiers: Convolutional Neural Network (CNN), Pearson correlation
• Feature extraction: CNN spatial filtering (8x1 kernel, 16 filters), CNN temporal filtering (1x32 kernel with dilation 2, 8 filters), CNN 2D convolution (5x5 kernel, 4 filters), sliding windows (250ms, 2ms stride)
• Frequency bands: analyzed=[0.1, 40.0] Hz
• Spatial filters: CNN 8x1 spatial convolution (16 filters)

Cross-Validation

• Method: sequential train/test split
• Evaluation type: offline classification, iterative calibration (1-6 blocks)

Performance (Original Study)

• Accuracy: 95.6%
• Itr: 67.49 bits/min
• Selection Time S: 1.5
• Cnn Training Time S: 15.0
• Burst 40 Accuracy: 94.2
• Mseq 100 Accuracy: 85.0

BCI Application

• Applications: reactive BCI
• Environment: controlled laboratory
• Online feedback: False

Tags

• Pathology: Healthy
• Modality: EEG
• Type: reactive BCI, c-VEP, visual evoked potentials

Documentation

• Description: Burst c-VEP based BCI study comparing novel burst code sequences to traditional m-sequences at two amplitude depths (100% and 40%) to optimize classification performance, minimize calibration data, and improve user experience
• DOI: 10.1016/j.neuroimage.2023.120446
• Associated paper DOI: 10.1016/j.neuroimage.2023.120446
• License: CC-BY-4.0
• Investigators: Kalou Cabrera Castillos, Simon Ladouce, Ludovic Darmet, Frédéric Dehais
• Senior author: Frédéric Dehais
• Contact: kalou.cabrera-castillos@isae-supaero.fr
• Institution: Institut Supérieur de l'Aéronautique et de l'Espace (ISAE-SUPAERO)
• Department: Human Factors and Neuroergonomics
• Address: 10 Av. Edouard Belin, Toulouse, 31400, France
• Country: FR
• Repository: Zenodo
• Data URL: https://zenodo.org/record/8255618
• Publication year: 2023
• Funding: AID (Powerbrain project), France; AXA Research Fund Chair for Neuroergonomics, France; Chair for Neuroadaptive Technology, Artificial and Natural Intelligence Toulouse Institute (ANITI), France
• Ethics approval: University of Toulouse ethics committee (CER approval number 2020-334); Declaration of Helsinki
• Acknowledgements: This work was funded by AID (Powerbrain project), France, the AXA Research Fund Chair for Neuroergonomics, France and Chair for Neuroadaptive Technology, Artificial and Natural Intelligence Toulouse Institute (ANITI), France.
• Keywords: Code-VEP, Reactive BCI, CNN, Amplitude depth reduction, Visual comfort

External Links

• Source: https://zenodo.org/record/8255618
• Github: https://github.com/neuroergoISAE/burst_codes

Abstract

The utilization of aperiodic flickering visual stimuli under the form of code-modulated Visual Evoked Potentials (c-VEP) represents a pivotal advancement in the field of reactive Brain–Computer Interface (rBCI). This study introduces Burst c-VEP, an innovative variant involving short bursts of aperiodic visual flashes at 2-4 flashes per second. The proposed burst c-VEP sequences exhibited higher accuracy (90.5%-95.6%) compared to m-sequence counterparts (71.4%-85.0%) with mean selection time of 1.5s. Reducing stimulus intensity to 40% amplitude depth only slightly decreased accuracy to 94.2% while substantially improving user experience. The collected dataset and CNN architecture implementation are shared through open-access repositories.

Methodology

Twelve healthy participants completed an offline 4-class c-VEP protocol using a factorial design. EEG was recorded at 500 Hz using BrainProducts LiveAmp 32-channel system. Participants focused on cued targets with factorial manipulation of pattern type (burst vs m-sequence) and amplitude depth (100% vs 40%). Visual stimuli were presented on a 60 Hz Dell monitor. Burst codes consisted of brief flashes (~50ms) with minimum 200ms inter-burst interval, while m-sequences used Fibonacci-type LFSR with segmented 132-frame subsequences. A CNN architecture with spatial (8x1, 16 filters), temporal (1x32, 8 filters), and 2D convolution (5x5, 4 filters) layers decoded EEG using 250ms sliding windows with 2ms stride. Calibration data ranged from 1-6 blocks (8.8-52.8s). Classification used sequential train/test splits with Pearson correlation for target selection. VEP analysis examined amplitude, latency, and inter-trial coherence. Statistical analyses used 2×2 repeated measures ANOVA.

References

Kalou Cabrera Castillos. (2023). 4-class code-VEP EEG data [Data set]. Zenodo.(dataset). DOI: https://doi.org/10.5281/zenodo.8255618

Kalou Cabrera Castillos, Simon Ladouce, Ludovic Darmet, Frédéric Dehais. Burst c-VEP Based BCI: Optimizing stimulus design for enhanced classification with minimal calibration data and improved user experience,NeuroImage,Volume 284, 2023,120446,ISSN 1053-8119 DOI: https://doi.org/10.1016/j.neuroimage.2023.120446

Notes

.. versionadded:: 1.1.0 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 1.5.0 (Mother of All BCI Benchmarks) https://github.com/NeuroTechX/moabb