EEG dataset and OpenBMI toolbox for three BCI paradigms: an investigation into BCI illiteracy

Abstract

Background: Electroencephalography (EEG)-based brain-computer interface (BCI) systems are mainly divided into three major paradigms: motor imagery (MI), event-related potential (ERP), and steady-state visually evoked potential (SSVEP). Here, we present a BCI dataset that includes the three major BCI paradigms with a large number of subjects over multiple sessions. In addition, information about the psychological and physiological conditions of BCI users was obtained using a questionnaire, and task-unrelated parameters such as resting state, artifacts, and electromyography of both arms were also recorded. We evaluated the decoding accuracies for the individual paradigms and determined performance variations across both subjects and sessions. Furthermore, we looked for more general, severe cases of BCI illiteracy than have been previously reported in the literature.

Results: Average decoding accuracies across all subjects and sessions were 71.1% (± 0.15), 96.7% (± 0.05), and 95.1% (± 0.09), and rates of BCI illiteracy were 53.7%, 11.1%, and 10.2% for MI, ERP, and SSVEP, respectively. Compared to the ERP and SSVEP paradigms, the MI paradigm exhibited large performance variations between both subjects and sessions. Furthermore, we found that 27.8% (15 out of 54) of users were universally BCI literate, i.e., they were able to proficiently perform all three paradigms. Interestingly, we found no universally illiterate BCI user, i.e., all participants were able to control at least one type of BCI system.

Conclusions: Our EEG dataset can be utilized for a wide range of BCI-related research questions. All methods for the data analysis in this study are supported with fully open-source scripts that can aid in every step of BCI technology. Furthermore, our results support previous but disjointed findings on the phenomenon of BCI illiteracy.

Keywords: BCI illiteracy; EEG datasets; OpenBMI toolbox; brain-computer interface; event-related potential; motor-imagery; steady-state visually evoked potential.

Figure 1:

The channel configuration of the International 10-20 system (62 EEG and 4 EMG recording electrodes). The left panel indicates the indexing; the right panel corresponding location of each electrode.

Figure 2:

Experimental designs for the three BCI paradigms. The 6×6 ERP speller layout (A), binary class MI (B), and four target frequencies SSVEP (C) paradigms were sequentially performed.

Figure 3:

Visualization of P300 responses (A), ERD/ERS patterns (B), and PSD (C) for ERP, MI, and SSVEP data, respectively. In the visualization of ERP (A) and MI (B) data, the first two rows show grid plots in time (x-axis) and amplitude (y-axis) domains for grand-averaged brain responses in certain channels (ERP: Cz and Oz, MI: C3 and C4). The next two rows indicate the topographies of entire brain area for each class corresponding to the certain time intervals that are displayed as gray areas in the grid plot. Fifth and sixth rows present topographic and grid plot, respectively, for signed r-values (significance level) between the binary classes. In the visualization of SSVEP data (C), one-dimensional data at Oz electrode were extracted and PSD was calculated in a frequency range of 0.1 to 25 Hz (x-axis).

Figure 4:

Average decoding accuracies in three BCI datasets over all subjects and sessions. The MI data were validated based on the CSP-cv, CSP, and more advanced algorithms (i.e., CSSP, FBCSP, and BSSFO). The decoding accuracies of ERP and SSVEP data were validated based on mean amplitude of ERP features and CCA, respectively.

Figure 5:

Scatter plots of performance variation across all subjects between sessions and paradigms. The first row shows variations of decoding accuracy in individual paradigms between sessions. Blue and gray circles indicate universally and partially literate BCI users, respectively, calculated in common decoding accuracy for the three BCI paradigms. The second row displays performance comparisons between paradigms (r, correlation coefficient).

Figure 6:

Mean rating scores of the questionnaire. Averages are calculated across all subjects and sessions. Four states such as concentration, eye-fatigue, and conditions of physical and mental state were representatively chosen (1 point: very low, 5: very high).

Figure 7:

Band power (dB) of resting state data in alpha frequency range (8–12 Hz). Twenty sets of resting state date, recorded during the entire experiment, were validated (see Table 3 for further information).