Home

This plugin removes low-frequency drifts, flatline channels, noisy channels, short-time bursts and incompletely repaired segments from the data. Hopefully parameter tuning should be the exception when using this function -- however, there are 3 parameters governing how aggressively bad channels, bursts, and irrecoverable time windows are being removed, plus several detail parameters that only need tuning under special circumstances.

Notes: This plugin uses the Signal Processing toolbox for pre- and post-processing of the data (removing drifts, channels and time windows); the core ASR method (clean_asr) does not require this toolbox but you will need high-pass filtered data if you use it directly.

Below we detail the GUI interface. Individual function contain additional help information.

Check checkbox (1) if the data has not been high pass filtered yet. If you use this option, the edit box in (2) allow setting the transition band for the high-pass filter in Hz. This is formatted as[transition-start, transition-end]. Default is 0.25 to 0.75 Hz.

Check checkbox (3) to reject bad channels. Options (4) allows to remove flat channels. The edit box sets the maximum tolerated flatline duration in seconds. If a channel has a longer flatline than this, it will be considered abnormal. The default is 5 seconds. Option (5) sets the Line Noise criterion: If a channel has more line noise relative to its signal than this value, in standard deviations based on the total channel population, it is considered abnormal. The default is 4 standard deviations. Option (6) sets the minimum channel correlation. If a channel is correlated at less than this value to an estimate based on other channels, it is considered abnormal in the given time window. This method requires that channel locations are available and roughly correct; otherwise a fallback criterion will be used. The default is a correlation of 0.8.

Check checkbox (4) to use Artifact Subspace Reconstruction (ASR). In edit box (5) you may change the standard deviation cutoff for removal of bursts (via ASR). Data portions whose variance is larger than this threshold relative to the calibration data are considered missing data and will be removed. The most aggressive value that can be used without losing much EEG is 3. For new users it is recommended to at first visually inspect the difference between the original and cleaned data to get a sense of the removed content at various levels. A quite conservative value is 10 and is the current default value. Use edit box (6) to use Riemannian distance instead of Euclidian distance. This is a beta option as the advantage of this method has not been demonstrated clearly yet. Checkbox (7) allow to remove instead of correcting the bad portions of data detected by ASR.