cleanline Cleanline

antillipsi@stage.nitrcce.org (Tim Mullen) — Thu, 09 Feb 2012 6:37:00 GMT

<pre class="plaintext" style="margin-top: 20px; margin-right: 0px; margin-bottom: 20px; margin-left: 0px; border-image: initial; font-size: 12px; font-family: Monaco, 'Courier New', monospace; vertical-align: baseline; line-height: 16px; border-top-left-radius: 3px; border-top-right-radius: 3px; border-bottom-right-radius: 3px; border-bottom-left-radius: 3px; background-image: initial; background-attachment: initial; background-origin: initial; background-clip: initial; white-space: pre-wrap; color: #393939; background-position: 0px 50%; padding: 0px; border: 0px initial initial;">Welcome to the CleanLine plugin for EEGLAB!

This plugin adaptively estimates and removes sinusoidal (e.g. line) noise from your ICA components
or scalp channels using multi-tapering and a Thompson F-statistic.

CleanLine is written by Tim Mullen ([url=mailto:tim@sccn.ucsd.edu]tim@sccn.ucsd.edu[/url]) with thanks to Makoto Miyakoshi for beta
testing. CleanLine makes use of functions modified from the Mitra Lab's Chronux
Toolbox ([url=http://www.chronux.org/]www.chronux.org[/url]).

CleanLine also makes use of the arg() functionality from Christian Kothe's BCILAB toolbox
(sccn.ucsd.edu/wiki/BCILAB)

----------------------------------------------------------------------------------------------------
INSTALLATION INSTRUCTIONS
----------------------------------------------------------------------------------------------------

Installation of CleanLine is simple:

1) download CleanLine (if you are reading this, you have probably already completed this step)
2) Unzip the package and copy to /plugins/. Start eeglab from the Matlab command line.
Alternately, you may add CleanLine (with subfolders) to your path and ensure EEGLAB is present
in the path.
3) If using the EEGLAB GUI you may start CleanLine from Tools-->CleanLine.
Alternately, you can start CleanLine from the command line
>> EEGclean = pop_cleanline(EEG);
See "Command-line example" section below for command-line example and other parameters
4) For Help, see Readme.txt or type
>> doc cleanline
Or, hold mouse over any textbox, checkbox, etc in the GUI for tooltip help text.

----------------------------------------------------------------------------------------------------
THEORY
----------------------------------------------------------------------------------------------------

Sinusoidal noise can be a prominent artifact in recorded electrophysiological data. This can stem
from AC power line fluctuations (e.g. 50/60 Hz line noise + harmonics), power suppliers (e.g. in
medical equipment), fluorescent lights, etc. Notch filtering is generally undesirable due to
creation of band-holes, and significant distortion of frequencies around the notch frequency (as well
as phase distortion at other frequencies and Gibbs rippling in the time-domain). Other approaches for
sinusoidal ("line") noise reduction include adaptive regressive filtering approaches (e.g. RLS, LMS),
but these typically require a reference signal (e.g. externally-recorded noise reference), which is
often unavailable. Blind-source separation approaches such as ICA may help mitigate line noise, but
often fail to completely remove the noise due to spatiotemporal non-stationarities in the noise.

CleanLine uses an approach for line noise removal advocated by Partha Mitra and Hemant Bokil in
"Observed Brain Dynamics" (2007), Chapter 7.3.4.

In brief, the data is traversed by a sliding window. Within each window, the signal is transformed to
the frequency domain using a multi-taper FFT. The complex amplitude (amplitude and phase) is thus
obtained for each frequency. Under the assumption of a deterministic sinusoid embedded in white
noise, we can set up a regression of the multi-taper transform (spectrum) of this sinusoidal signal
onto the multitaper spectrum of the original data at a given frequency. The regression coefficient
is a complex number representing the complex amplitude (phase and amplitude) of the deterministic
sinusoid. From this, a time-domain representation of the sinusoid may be constructed and subtracted
from the data to remove the line.

Typically, one does not know the exact line frequency. For instance, in the U.S.A., power line noise
is not guaranteed to be at exactly 60 Hz (or even to have constant phase over a given period of time).
To ameliorate this problem a Thompson F-Test may be applied to determine the statistical significance
of a non-zero coefficient in the above regression (indicating a sinusoid with significantly non-zero
amplitude). We can then search within a narrow band around the expected location of the line for the
frequency which maximizes this F-statistic above a significance threshold (e.g. p=0.05).

Line frequency scanning can be enabled/disabled using the 'ScanForLines' option.

Overlapping short (e.g. 2-4 second) windows can be used to adaptively estimate the frequency, phase,
and amplitude of the sinusoidal noise components (which typically change over the course of a recording
session). The discontinuity at the point of window overlap can be smoothed using a sigmoidal function.</pre><pre class="plaintext" style="margin-top: 20px; margin-right: 0px; margin-bottom: 20px; margin-left: 0px; border-image: initial; font-size: 12px; font-family: Monaco, 'Courier New', monospace; vertical-align: baseline; line-height: 16px; border-top-left-radius: 3px; border-top-right-radius: 3px; border-bottom-right-radius: 3px; border-bottom-left-radius: 3px; background-image: initial; background-attachment: initial; background-origin: initial; background-clip: initial; white-space: pre-wrap; color: #393939; background-position: 0px 50%; padding: 0px; border: 0px initial initial;">The smoothing factor is determined by the 'SmoothingFactor' parameter in cleanline().

CleanLine allows you to specify the multi-taper frequency resolution by the 'Bandwidth' parameter.
This is the width of a peak in the spectrum for a sinusoid at given frequency. Due to the time-frequency
uncertainty principle, decreasing bandwidth increases the necessary length of the sliding window in
order to obtain a reasonable frequency decomposition. The number of tapers, K, used by CleanLine is
given by K=2TW-1 where T is the temporal resolution in seconds (window length) and W is the frequency
resolution (Bandwidth) in Hz. CleanLine fixes T to be the sliding window length, so W (Bandwidth) is the
only required parameter. If the 'Verbosity' option is set to 1, then CleanLine will display the multi-taper
parameters in the command line on execution of the function.

----------------------------------------------------------------------------------------------------
TIPS ON RUNNING CLEANLINE
----------------------------------------------------------------------------------------------------

The default options should work quite well, but parameters may need to be tweaked depending on the setup.
If you have multiple epochs you need to make sure that your window size and step size exactly divides the
epoch length. In other words, you do not want any sliding windows to overlap two epochs since line noise
phase and amplitude may shift at that point, making it impossible to perform the time-domain line subtraction.
If you have relatively short epochs (e.g. < 5 sec) it is best if each window is taken to be the length
of the epoch and the step size is equal to the window length. In this way, the lines are estimated and
removed for each epoch individually. When using the GUI, the default values for window length and step size
are automatically set to the epoch length.

If cleaning continuous, un-epoched data, then you may wish to use sliding windows of 3-4 seconds with 50%
overlap.

If using the EEGLAB GUI, commands are stored in EEG.history so that the eegh() command will return the
command-line function call corresponding to the last GUI execution of CleanLine.

You might find it useful to try the option ('PlotFigures',true) on a subset of channels/components to get a
sense of the performance of CleanLine for difference parameter choices (and also to identify where the most
significant lines lie in the spectrum) and then, once you are satisfied with the parameters, turn this option
off before cleaning all the remaining channels/components. NOTE: CleanLine is *considerably slower* if PlotFigures
is enabled. If you don't care to see the visualize the final results of the cleaning operation, you may also
wish to set ('ComputeSpectralPower',false) which will speed up computation considerably.
</pre>

CleanLine Releases

cleanline Cleanline