Table 10.
Signal processing and artifact removal methods for EEG studies.
| Name, year | Pre-processing | Spatial filtering | Movement artifact removal |
|---|---|---|---|
| Adkin et al., 2008 | Bandpass filter (0.1–10 kHz) | ||
| Low-pass filter (30 Hz cutoff) | |||
| Adkin et al., 2006 | Bandpass filter (0.0001–30 Hz) | Manually removed trials with artifact | |
| Beurskens et al., 2016 | Bandpass filter (0.5–45 Hz) | Reference using common reference | Visual inspection and semiautomatic artifact rejection (±100 μV) |
| Bradford et al., 2015 | Noisy channels removed | AMICA and DIPFIT | |
| Bruijn et al., 2015 | High-pass filter (3 Hz Butterworth) | Average common reference | ICA |
| Band-stop filter (50, 100, 150, and 250 Hz) | |||
| Bulea et al., 2014 | High pass filter (Butterworth, 0.05 Hz) | Common average reference | Artifact subspace reconstruction (uses PCA to clean data) |
| Bandpass filter (Butterworthy, 0.1–4 Hz) | |||
| Bulea et al., 2015 | High pass filter at 1 Hz | Re-referenced to common average of the remaining channels | Artifact subspace reconstruction (uses PCA to clean data). AMICA and DIPFIT |
| Power line noise remove | |||
| Noisy channels removed | |||
| Chang et al., 2016 | Bandpass filter (70 Hz–DC) | ||
| Notch filter (60 Hz) | |||
| Bandpass filter at (0.1–50 Hz) | |||
| De Sanctis et al., 2014 | Bandpass filter (0.05–100 Hz) | Re-referenced offline to an average reference | Automatic artifact rejection (±100 μV) |
| Bandpass filter (0.5–30 Hz) | |||
| Del Percio et al., 2007, 2009 | Bandpass filter (0.01–100 Hz) | Laplacian estimation | Autoregressive method to remove ocular artifact |
| Gramann et al., 2010 | High pass filter (1 Hz) | Re-referenced offline to an average reference | ICA, AMICA and DIPFIT |
| Noisy channels removed | |||
| Gwin et al., 2010 | High pass filter (1 Hz) | Re-referenced offline to an average reference | Moving average method |
| Noisy channels removed | |||
| Gwin et al., 2011 | High pass filter (1 Hz) | Re-referenced offline to an average reference | Rejected time periods with substantial artifact. AMICA and DIPFIT |
| Noisy channels removed | |||
| Haefeli et al., 2011 | Bandpass filter (1–30 Hz) | Average reference as recording reference | All artifacts exceeding ± 80 μV excluded, then ICA |
| Huang et al., 2014 | AC amplifier with cutoff freq 5–450 Hz | Visually inspected and removed artifacts. PCA prior to ERP spectral analysis | |
| Bandpass filter (100 Hz) | |||
| Low pass filter (40 Hz) | |||
| Hülsdünker et al., 2015, 2016 | Band pass filter (2–120 Hz) | Re-referenced to a common average reference. | Semiautomatic rejection algorithm followed by ICA |
| Jacobs et al., 2008 | Bandpass filter (0.05–60 Hz) | ||
| Kline et al., 2014 | High pass filter (1 Hz) | Re-referenced the remaining channels to an average reference | AMICA and DIPFIT |
| Noisy channels removed | |||
| Kline et al., 2015 | High pass filter (1 Hz) | Re-referenced the remaining channels to an average reference. | Moving average method |
| Wavelets | |||
| Rejected epochs > 3 standard deviations from the means of the gait event times | Moving Average and Wavelets | ||
| EEG movement artifact compared to accelerometer signal | |||
| Lau et al., 2014 | Highpass filter (1 Hz). | Re-referenced the remaining channels to an average reference | Moving average method |
| Filter line noise (60 Hz) | |||
| Noisy channels removed | |||
| Lau et al., 2012 | Highpass filter (1 Hz). | Channels were then re-referenced to an average of the remaining channels. | Weighted Phase Lag Index |
| Filter around 4+/−2H | |||
| Noisy channels removed | |||
| Little and Woollacott, 2015 | Notch filter (60 Hz) | Artifact detection algorithm and visual inspection | |
| High pass filter (30 Hz) | |||
| Low pass filter (0.1 Hz) | |||
| Luu et al., 2016 | Adaptive filter | Artifact subspace reconstruction (uses PCA to clean data) | |
| Malcolm et al., 2015 | Low pass filter (Butterworth 7 Hz); Bandpass filter (1–30 Hz) | Re-referenced to an average reference | Artifact rejection criterion (±75 μV) |
| Marlin et al., 2014 | Filter (DC–300 Hz) | Visual inspection to remove ocular artifact. PCA and ICA | |
| Low pass filter (30 Hz) | |||
| Mierau et al., 2015 | Butterworth IIR filter,Bandpass filter (2–30 Hz) | Laplacian interpolation | Corrected for ocular artifacts |
| Mochizuki et al., 2008, 2009a,b | Low pass (200 Hz) | Removal of ocular artifact | |
| High pass (DC) | |||
| Low pass (30 Hz) | |||
| Nathan and Contreras-Vidal, 2016 | Low pass (DC–1,000 Hz) | Artifact subspace reconstruction (uses PCA to clean data) | |
| Oliveira et al., 2016 | High pass filter (1 Hz) | Removed frame sequences with large artifacts due to lost packets during wireless telemetry and EMG. ICA | |
| Notch filter using Cleanline | |||
| Petersen et al., 2012 | Filter (1–500 Hz) | Re-referenced to a common average reference | ICA |
| Removed significant drift or >50 Hz noise | |||
| Petrofsky and Khowailed, 2014 | Bandpass filter (0.1–65Hz) | Mean wavelet coefficients per frequency bin. Linear discriminant function analysis to classify eye blinks | |
| Notch filter (60z) | |||
| Removed amplitude saturation | |||
| Pirini et al., 2011 | Manual cleaning and ICA | ||
| Presacco et al., 2011, 2012 | Bandpass filter (0.1–100 Hz) | None | |
| Bandpass filter (Butterworth 0.1–2 Hz) | |||
| Quant et al., 2004 | Bandpass filter (1–10,000 Hz) | Visually inspected for artifact and averaged over each subject | |
| Quant et al., 2005 | Bandpass filter (1–10,000 Hz); Butterworth low-pass filter (30 Hz) | Visually inspected for artifact and averaged over each subject | |
| Sipp et al., 2013 | Bandpass filter (DC–104 Hz) | Remaining channels were average referenced. | Moving average method |
| High pass filter (1 Hz) | |||
| Noisy channels removed | |||
| Slobounov et al., 2015 | Filter (2–30 Hz) | Manual check for artifacts and eye blinks | |
| Slobounov et al., 2013 | Filter (4–30 Hz) | Checked and corrected for artifacts and eye blinks removed | |
| Slobounov et al., 2009 | Filter (70 Hz) | ICA, Morlet wavelet transformation | |
| Slobounov et al., 2008 | Filter (100 Hz) | Visual inspection | |
| Slobounov et al., 2005 | Filter (100 Hz) | Ocular artifact reduction through NeuroScan software | |
| Smith et al., 2012, 2014 | Low pass filter (40 Hz) | Re-referenced to a common average reference | ICA |
| Snyder et al., 2015 | Low pass filter (1 Hz) | Re-referenced to average | AMICA followed by DIPFIT |
| Noisy channels removed | |||
| Storzer et al., 2016 | Bandpass filter (1–100 Hz) | Re-referenced against the common average | Visual inspection then ICA |
| Bandstop (49–51 Hz) | |||
| Tse et al., 2013 | Bandpass filter (0.1–65 Hz) | Mean wavelet coefficients per frequency bin. Linear discriminant function analysis to classify eye blinks | |
| Notch filter (60 Hz) | |||
| Varghese et al., 2016 | Bandpass filter (0.05–50 Hz) | ICA | |
| Varghese et al., 2015 | Bandpass filter (DC–300 Hz) | ICA | |
| Bandpass filter (2–50 Hz) | |||
| Varghese et al., 2014 | Bandpass filter (1–30 Hz) | ICA |
ICA, Independent component analysis; AMICA, Adaptive mixture ICA; PCA, Principal component analysis; DIPFIT, dipole source localization of independent components; ERP, event related potential.