Table 6 The table shows EEG recording apparatus and EEG processing software
Reference | EEG apparatus | Software(Bold) and processing | |||
|---|---|---|---|---|---|
EEG Cap | On-line | Off-line | On-line | Off-line | |
Alatorre et al. [61] | ElectroCap™ (International Inc.; Eaton, Ohio) | MEDICID IV System. Sampling rate at 200 Hz. | MEDICID IV System. Sampling rate at 200 Hz | Track Walker: Theta absolute power was segmented into a 1280ms-window, displacing 20ms each time. z-value for theta AP was computed | A custom software [94, 95]: statistical nonparametric mapping(SnPM) was performed. Artefact-free segments of 2.56s were selected. Cross-spectral matrices and absolute EEG power was computed by fast Fourier Transform (FFT). Relative power of interest was Z-transformed for NF |
Andrade et al. [58] | 20-channel cap | Neuroelectrics, Barcelona, Spain; Sampling rate at 500 Hz | NIC® EEG amplifier, Neuroelectrics | MATLAB Psychtoolbox v3 and System Level Simulations (SLS): no method reported | MATLAB Brainstorm: gamma, peak alpha and theta/beta ratio were computed using Welch’s PSD estimation on 1 s-window with 25% overlapping in each 2-seconds epoch |
Azarpaikan and Torbati [47] | 20-channel cap | Flex Comp Infiniti encoder and TT-USB interface unit. Sampling rate at 256 Hz | Flex Comp Infiniti encoder. Sampling rate at 256 Hz | BioGraph Infiniti Software System: no method reported | Brain Vision Analyzer: Absolute SMR and theta power were elicited by complex demodulation |
Becerra et al. [60] | 19-lead cap | MEDICID IV System | MEDICID IV System | MEDICID IV: Band power was computed by FFT. Data was Z-transformed. 20 ms of absolute band power was computed every 5 ms for feedback | MEDICID IV: Absolute and relative power of feedback frequencies were computed by FFT for analysis |
Campos et al. [56] | 19-channels, (Waveguard connect) | ProComp Infiniti differential amplifier (Thought Technology Ltd) | NeuronSpectrum-4/EP system | BioGraph Infiniti software: Spectral amplitudes were computed on raw 1s EEG segments | MATLAB EEGLAB toolbox: EEG data were digitised into non-overlapping epochs time-locked to the task condition. The eye muscle artefacts were rejected by independent component analysis (ICA) |
Gomez et al. [59] | 8-lead elastic cap | g.USBamp amplifier (Guger Technologies). Sampling rate at 256 Hz. | g.USBamp amplifier (Guger Technologies). Sampling rate at 256 Hz. | BCI2000: Spatial nearest - neighbour Laplacian was applied over feedback electrodes. | Matlab R2011b: The artefacts were rejected using ICA. Band power spectral density was computed using nonparametric Welch method with 32s Hamming window. |
Jirayu et al. [46] | 14-channel Emotiv cap | Emotiv EPOC headset. Sampling rate at 128 Hz | Emotiv EPOC headset. Sampling rate at 128 Hz | MATLAB: ICA with Lifting Wavelet transform (LWT) to improve artefact removal efficiency. Power spectrum was computed by FFT with a Hanning Window sized 256 samples. Moving average of attention (ratio of Beta/Alpha) value in the window of 2 seconds was calculated. Each processed data block size is assigned to 64 samples | No processing software or method reported |
Kober et al. [53] | 10-channel cap | NeXus-10 MKII, Mind Media BV. Sampling rate at 256 Hz | Not clear | BioTrace+(Mind Media): Power spectra was computed using FFT, with 10\(\%\) Hannning window | Brain Vision Analyzer: Frequency powers were calculated using the built-in method of complex demodulation. The data was Z-transformed. All epochs with artefacts were excluded from the analysis |
Kober et al. [51] | 10-channel cap | NeXus-10 MKII, Mind Media BV. Sampling rate at 256 Hz | Not reported | BioTrace+(Mind Media): Power spectra was computed using FFT, with 10\(\%\) Hannning window | Brain Vision Analyzer: Frequency powers were calculated using the built-in method of complex demodulation. The data was Z-transformed. All epochs with artefacts were excluded from the analysis |
Lee et al. [49] | Self-developed EEG headband with 2 dry electrodes | Recording system not reported | Off-line EEG not recorded | Custom platform: Spatial pattern filtering computed a model for discriminating attentive and inattentive states. A classifier in the model transformed the patterns into attention level for NF. EEG data was quantified into attention score at every 200 ms | No processing software or method reported |
Marlats et al. [55] | Electro Cap (International Inc. Elmiko Medical) | EEG Digitrack Biofeedback plus module (Inc Elmiko Medical). Sampling rate at 512Hz | EEGDigitrack SimplEEG32. Sampling rate at 500 Hz | EEG Digitrack Biofeedback plus module (Inc Elmiko Medical): Moving averaged EEG rhythms (delta, theta, alpha, SMR, and lower beta) were computed | Python and R: ICA was used to reject artefacts and the recorded EEG passed a rejection test with a covariance-based approach. Relative band power was computed by FFT and averaged with Welch method |
Reichert et al. [52] | 10-channel cap | 10-channel system NeXus-10 MKII, Mind Media. Sampling rate at 256 Hz | 16-channel g.USBamp (g.tec). Sampling rate at 500 Hz | Brain Vision Analyzer(Brain Products GmbH): Band power was computed using built-in complex demodulation | Brain Vision Analyzer software (Brain Products GmbH): The data was Z-transformed. All epochs with artefacts were excluded from the analysis |
Reis et al. [54] | actiCAP and EASYcCAP (Brain Product) | QuickAmp (Brain Products) or ActiCHamp (Brain Products). Sampling rate at 500 Hz | QuickAmp (Brain Products) or ActiCHamp (Brain Products). Sampling rate at 500 Hz | Custom platform in BCI++ platform (developed by C++ and Matlab): No artefact rejection was performed. Relative power of interest was calculated within a 1s-window and update visually to the user every 200ms | EEGLAB by MATLAB and OriginLab: Two-way least square FIR filter was applied to discard muscle artefacts and equipment noise. Ocular artefacts were removed based on ICA. Frequency bands absolute and relative PSD were computed |
Sanchez-Cifo et al. [48] | 2-channel EEG device Muse 2 | Muse 2. Sampling rate at 256 Hz | / | Python and MATLAB: Frequency bands were computed to train Support Vector Machine (SVM) for emotional state classification | No processing software or method reported |
Staufenbiel et al. [57] | EEG cap (g.tec medical engineering GmbH) | USB Biosignal Amplifier (g.tec medical engineering GmbH). Sampling rate at 256 Hz | USB Biosignal Amplifier (g.tec medical engineering GmbH). Sampling rate at 256 Hz | Custom-built software: Band power was computed by FFT with 1-second delay. Band power level was based on a moving average of 30s updated continuously. Average power was calculated over epochs per second | Brain Vision Analyzer 2.0 (BrainProducts): Data was subjected to ocular correction. All epochs (200ms-window) with artefacts were rejected |
Wang and Hsieh [50] | Neuroscan Q-cap AgCl-32 electrode cap | ProComp Infiniti differential amp and software. Sampling rate at 256 Hz | Synamps2 (Neuroscan). Sampling rate at 500Hz | BioGraph Infiniti software: IIR filter was applied to extract frequency-specific bands. EEG signals were segmented into 1 Hz with 1-second time window. Reward EEG frequency bands were extracted through FFT | Scan 4.3 acquisition software (Neuroscan): To acquire artefact-free data, ocular artefacts reduction method was employed. Absolute EEG power was computed by FFT |