Table 1.
Important fNIRS-BCI studies from 2004 to 2014.
| References | Brain area | Task | Filter used for noise removal | Features | Classifier | Classification accuracy (%) |
|---|---|---|---|---|---|---|
| Coyle et al., 2004 | Motor cortex | Motor imagery | Low-pass | Mean values of ΔHbO | Threshold-based | 75 |
| Sitaram et al., 2007 | Motor cortex | Motor imagery | Low-pass | Mean of ΔHbO and ΔHbR for all channels | SVM and HMM | 73 (SVM) |
| 89 (HMM) | ||||||
| Naito et al., 2007 | Prefrontal cortex | Mental arithmetic, music imagery and landscape imagery | Low-pass | Amplitude of light intensity | QDA | 80 |
| Coyle et al., 2007 | Motor cortex | Motor imagery | Low-pass | Mean values of ΔHbO | Threshold-based | 80 |
| Luu and Chau, 2009 | Prefrontal cortex | Neural correlates of subjective preference | Low-pass | Mean amplitude of light-intensity signals | LDA | 80 |
| Tai and Chau, 2009 | Prefrontal cortex | Emotion rehearsal associated with images | Least-mean square adaptive filter | Mean, variance, zero crossing, root mean square, skewness and kurtosis of ΔHbO and ΔHbR | LDA and SVM | 96.6 (LDA) |
| 94.6 (SVM) | ||||||
| Power et al., 2010 | Prefrontal cortex | Mental arithmetic/Music imagery | Low-pass/Wavelet filter | Mean values of light intensity | HMM | 77.2 |
| Cui et al., 2010b | Motor cortex | Finger tapping | Exponential moving average filter | Amplitude, history, history gradient and 2nd order gradient of ΔHbO and ΔHbR, spatial patterns | SVM | >80 (spatial patterns) |
| Abibullaev et al., 2011 | Prefrontal cortex | Object rotation, verbal fluency and mental arithmetic | Wavelet filter | Mean, power, standard deviation etc. of the filter coefficients from wavelet transform | ANN | > 94 |
| Falk et al., 2011* | Prefrontal cortex | Music imagery | Low-pass | ΔHbO and ΔHbR values after wavelet transform | HMM | 83 |
| Holper and Wolf, 2011 | Motor cortex | Motor imagery | Low-pass | Mean, variance, skewness and kurtosis of ΔHbO | LDA | 81 |
| Tanaka and Katura, 2011 | Prefrontal and Visual cortex | Change-detection task | Moving average/ Band-select | ΔHbO and ΔHbR values from single, two and three channels | SVM | 77 |
| Bauernfeind et al., 2011 | Prefrontal cortex | Mental arithmetic | High-pass | Antagonistic ΔHbO signals | LDA | 79.7 |
| Power et al., 2011 | Prefrontal cortex | Mental arithmetic/Mental singing | Low-pass | Slope of light-intensity signals | LDA | 71.2 (mental arithmetic) |
| 62.7 (mental singing) | ||||||
| Chan et al., 2012 | Prefrontal cortex | Mental singing | Low-pass | ΔHbO and ΔHbR signals from selected channels | ANN and HMM | 63 (ANN) |
| 55.7 (HMM) | ||||||
| Fazli et al., 2012** | Motor cortex | Motor execution and motor imagery | Low-pass | Mean values of ΔHbO and ΔHbR and band power of Laplacian filtered EEG data | LDA | 92.6 (EEG+HbR for motor execution) |
| 83.1 (EEG+HbR for motor imagery) | ||||||
| Seo et al., 2012 | Motor cortex | Finger tapping | Moving average/band-pass | Raw ΔHbO values at three time points | PLSDA and HMM | 90.2 (PLSDA) |
| 85.7 (HMM) | ||||||
| Hu et al., 2012 | Prefrontal cortex | Deception | Band-pass | Absolute values of ΔHbO and ΔHbR | SVM | 83.4 |
| Power et al., 2012b | Prefrontal cortex | Mental arithmetic | Low-pass | Signal slope | LDA | 72.6 |
| Moghimi et al., 2012 | Prefrontal cortex | Emotionally rated music listening | Low-pass | Mean and difference between signal and noise of ΔHbO and ΔHbR | LDA | 71.9 |
| Abibullaev and An, 2012 | Prefrontal cortex | Object rotation, letter padding and multiplication | Wavelet filter | Filter coefficients from wavelet transform | ANN, LDA and SVM | >75 (ANN) |
| >85 (LDA) | ||||||
| >90 (SVM) | ||||||
| Liu et al., 2013 | Prefrontal cortex | Neural correlation of visual stimulus | Band-pass | Mean values of ΔHbO and ΔHbR and EEG amplitudes | Step-wise LDA | >90 |
| Power and Chau, 2013 | Prefrontal cortex | Mental arithmetic | Low-pass | Signal slope of ΔHbO and ΔHbR | LDA | 71.1 |
| Stangl et al., 2013 | Motor cortex, prefrontal cortex | Motor imagery, mental arithmetic | Moving average | Amplitude of ΔHbO | LDA | 65 |
| Zimmermann et al., 2013* | Motor cortex | Motor execution | Band-pass | Combination of ΔHbO and ΔHbR and biosignals belonging to same location and same time | HMM | 88.5 |
| Naseer and Hong, 2013 | Motor cortex | Motor imagery | Low-pass | Mean and slope values of ΔHbO and ΔHbR in several temporal windows | LDA | 77.5 (mean value) |
| 87.2 (signal slope) | ||||||
| Faress and Chau, 2013*** | Prefrontal cortex | Verbal fluency | Low-pass | Slope of HbO, HbR and HbT | LDA | 86 |
| Hai et al., 2013 | Motor cortex | Hand tapping | Savitzky-Golay | Signal values after polynomial regression | SVM and ANN | 79.1 (SVM) |
| 83.3 (ANN) | ||||||
| Schudlo and Chau, 2014 | Prefrontal cortex | Mental arithmetic | Low-pass | Slope of ΔHbO, ΔHbR and ΔHbT in 3 different time windows | LDA | 77.4 |
| Naseer et al., 2014 | Prefrontal cortex | Mental arithmetic | Low-pass | Mean values of ΔHbO and ΔHbR | LDA and SVM | 74.2 (LDA) |
| 82.1 (SVM) | ||||||
| Khan et al., 2014** | Motor cortex, prefrontal cortex | Motor execution, mental counting and mental arithmetic | Band-pass | Mean values of ΔHbO and ΔHbR | LDA | >80 |
| Shin and Jeong, 2014 | Motor cortex | Motor execution | Band-pass and Savitzky-Golay | Mean, amplitude, slope, delay, variance and median | Naïve Bayes classifier | 95.5 (binary) |
| 92.4 (ternary) | ||||||
| 91.5 (quaternary) | ||||||
| Hwang et al., 2014 | Motor cortex, Prefrontal cortex | Motor Imagery, mental singing, mental arithmetic, mental rotation and mental character writing | Band-pass | Mean values of HbO, HbR and HbT | LDA | >70 (mental arithmetic and mental rotation) |
| Hong et al., 2015 | Motor cortex, prefrontal cortex | Motor imagery, mental arithmetic | Band-pass | Mean and slope of HbO | Multi-class LDA | > 75 |
*
For classification purpose, they also used some additional features from ECG, respiration, blood pressure, skin conductance response, etc.
**
They used combined fNIRS and EEG modalities for brain signal acquisition.
***
This study used additional transcranial Doppler ultrasonography signals together with fNIRS signals.