Table 2.
Online signal-processing methods and hardware.
| Study | Data for section 2.1 | Data for section 2.2 | Data for section 2.4 | Data for section 2.3 | Data for section 2.5 | ||
|---|---|---|---|---|---|---|---|
| Device | Selections of target regions [channel(s) of interest] | Online feature | Chromophore | Online preprocessing | Artifact control | Calculation of feedback signal | |
| Aranyi et al. (2016) | fNIR400 | 8 channels positioned on subject's forehead | Asymmetry (left vs. right dlPFC amplitudes) | HbO | Low-pass filter (finite impulse response, order 20) 0.1 Hz, sliding-window motion artifact rejection, reference channel | Reference channels | Threshold of signal based on mean, SD, and signal variation of signal during counting (baseline) |
| Barth et al. (2016) | ETG-4000 | 14 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki et al., 2007) | Amplitude, PFC | HbO | Not reported | None | Change in HbO compared to 15 s baseline at beginning. |
| Fujimoto et al. (2017) | OMM-3000 | 4 channels aligned with positions of EEG 10–20 system. MNI positions estimated using individual structural MRI and digitizer measurements | Amplitude (t-values estimated by GLM), SMA | HbO | GLM analysis. 20 s sliding window. Linear term to correct drift. 4 short-distance channels (principal component included as regressor in GLM). | EMG, 4 short-distance channels | Contrast regulation vs. rest (adaptive GLM, 20 s sliding window), maximum t-value from the 4 channels. Primary principal component of short-distance channels as nuisance regressor |
| Hosseini et al. (2016) | ETG-4000 | Channels aligned with positions of EEG 10–20 system (Okamoto et al., 2004). Functional localizer (working memory task) | Amplitude, dlPFC | HbO | Bandpass filter 0.01–0.5 Hz | No | Change in the average HbO signal over feedback channels (over 9 s window) relative to the calibration period |
| Hudak et al. (2017) | ETG-4000 | 8 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki et al., 2007) | Amplitude, dlPFC/IFG | HbO | Kalman filter with a 5 s sliding window, CAR of all channels | CAR | Change in HbO compared to 5 s baseline, CAR of all channels subtracted |
| Hudak et al. (2018) | ETG-4000 | 8 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki et al., 2007) | Amplitude, dlPFC/IFG | HbO | Bandpass filter: 0.01–0.1 Hz, 5 s moving average, CAR of all channels | CAR | Change in HbO compared to 5 s baseline, CAR of all channels subtracted |
| Kanoh et al. (2011) | ETG-4000 | 3 channels aligned with positions of EEG 10–20 system | Amplitude sensorimotor cortex | HbO | High-pass filter and 7-point moving average | None | Average of channels/no baseline before task (not explicitly reported) |
| Kimmig et al. (2018) | ETG-4000 | 10 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki and Dan, 2014) | Amplitude, dlPFC/IFG | HbO | 5 s moving average filter, CAR of remaining channels | CAR | Average of channels, change in HbO compared to 5 s baseline. Standard deviation from previous trial used to scale maximum/minimum of feedback signal. CAR of remaining channels subtracted. |
| Kinoshita et al. (2016) | ETG-4000 | 6 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki et al., 2007). | Amplitude, frontopolar cortex | HbO | Not reported | Respiratory rate | Average of channels, moving baseline, last 10 s. Maximum display of the bar graph was +0.25 [mMmm] |
| Kober et al. (2014) | ETG-4000 | 8 channels aligned with positions of EEG 10–20 system registered to MNI space (Singh et al., 2005). Functional localizer to select channels with best signal quality during a motor task | Asymmetry (difference between left and right motor area) | HbO | 0.01 HPF and 1.5 Hz LPF, 2 s moving average. Difference of right and left channels (also cancels out artifacts) | 4 reference channels | HbO left vs. HbO right. No baseline period before trials |
| Kober et al. (2015) | ETG-4000 | 4 channels aligned with positions of EEG 10–20 system, positions assessed via individual digitizer measurements | Amplitude (difference between IFG and posterior regions) | HbO and HbR group | 0.01 high- and 1.5 Hz low-pass filter, 2 s moving average. Difference of IFG and of posterior reference channels | EMG, 4 reference channels | HbO/HbR of FB channels vs. reference channels. No baseline period before trials |
| Kober et al. (2018) | NIRSport | 4 channels, positions assessed via individual digitizer measurements, probably based on EEG 10–20 system | Amplitude, IFG | HbO and HbR group | Not reported | None | Not reported |
| Lapborisuth et al. (2017) | LABNIRS | 14 channels aligned with positions of EEG 10–20 system, positions assessed via individual digitizer measurements | Amplitude, motor cortex | HbO | Detrending and normalization (last 10 s) | None | Difference of current sample divided by SD of previous 10 s and linear trend of previous 10 s |
| Lee et al. (2015) | FOIRE-3000 | 7 channels aligned with positions of EEG 10–20 system. Positions assessed via individual digitizer measurements | Amplitude, sensorimotor cortex | HbO | Not reported | None | t-values, probably GLM |
| Li et al. (2019) | NIRSport | 1 channel aligned with positions of EEG 10–20 system. Position of channel validated by MRI scans in two independent participants | Amplitude, OFC | HbO | 2 s moving average | None | Change in HbO compared to 2 s baseline, feedback scaled based on a pre-experiment (“difficulty coefficient”) |
| Liu et al. (2016) | ETG-4000 | Functional localizer at beginning of each session (channel with highest/lowest signal during face processing), positions assessed via individual digitizer measurements | Amplitude, frontal and temporal face processing regions | HbO | Reference channel (channel irrelevant for face-processing network identified during functional localizer) | Reference channel | HbO Percent signal change in feedback channel compared to reference channel (face matching compared to house matching) |
| Marx et al. (2015) | ETG-4000 | 8 channels aligned with positions of EEG 10–20 system, registered to MNI space (Tsuzuki et al., 2007) | Amplitude, dlPFC/IFG | HbO | Common average reference (CAR) | No | Change in HbO compared to 5 s baseline |
| Mihara et al. (2012) | OMM-3000 | 3 channels aligned with positions of EEG 10–20 system, positions estimated using structural MRI and digitizer measurements of representative participants | Amplitude (t-values estimated by GLM), premotor cortex | HbO | GLM analysis. 20 s sliding window, linear term to correct for drift. Autoregressive model order 1 to adjust autocorrelation, excluded 3 participants with finger movement | No, but reported offline control | Contrast regulation vs. rest (adaptive GLM, 20 s sliding window), maximum t-value from the 3 channels |
| Mihara et al. (2013) | OMM-3000 | 3 channels aligned with positions of EEG 10–20 system, positions estimated using individual MRI and digitizer measurements | Amplitude (t-values estimated by GLM), premotor cortex | HbO | GLM analysis, 20 s sliding window. Linear term to correct for drift, autoregressive model order 1 to adjust autocorrelation, EMG control | EMG | Contrast regulation vs. rest (adaptive GLM, 20 s sliding window), maximum t-value from the 3 channels |
| Narita (2015) | PocketNIRS, Dynasense | 1 channel aligned with positions of EEG 10–20 system. No information about registration reported. | Amplitude, PFC | HbO | Not reported | None | Not reported |
| Trambaiolli et al. (2018) | NIRScout | 32 channels aligned with positions of EEG 10–20 system. No information about registration reported | Output of classifier, frontal/occipital networks | HbO, HbR | 1 Hz low-pass filter and 3 s moving average, normalization by average signal from the same channel during previous neutral condition | None | Output of linear discriminant classifier based on HbO and HbR of 32 channels, indicating neutral or positive affect |
| Weyand et al. (2015) | Imagent | 9 channels aligned with positions of EEG 10–20 system. No information about registration reported | Classifier, PFC | HbO, HbR, tHb | Third-order Chebyshev IIR cascade filter with pass-band edge frequency of 0.1 Hz, stop-band edge frequency of 0.5 Hz, and pass band ripple of 0.1 dB | None | Session 1–5: weighted slope score. Session 6–16: output of classifier. LDA classifier using 8 features selected from 288 different temporal and spatiotemporal features from all channels and chromophores |
CAR, common average reference; COI, channel(s) of interest; dlPFC, dorsolateral prefrontal cortex; EEG, electroencephalography; EMG, electromyography; IFG, inferior frontal gyrus; HbR, deoxyhemoglobin; LDA, linear discriminant analysis; HbO, oxyhemoglobin; PFC, prefrontal cortex; tHb total hemoglobin.