[Eeglablist] artifact detection in continuous data

[Tarik S Bel-Bahar]

Greetings James,

Thanks for your always useful responses on the list!

There are multiple issues with artifact detection as you know. Groups with
large samples and not enough staffing will thank you for
such an automated process. Manousos and Mahesh, if this brings up any
thoughts or issues or sample code,
please let us know.

A few quick thoughts (kind of a wishlist really):

0. I'd be willing to be a beta tester, and provide data, and "artifact
detection"-by human data samples,
for the process. At minimum, the easiest first solution seemed to be (to
me), to
epoch-based artifact detection tools of your choice in EEGLAB, and apply
them in several sweeps,
to "artificial epochs" in your continous data.By artificial, i mean just
making contiguous 200 or 500 or 1000 ms epochs,
regrardless of events. This I sometimes do to clean up ICA decompostions
before applying them back to continuous data.


1. Arno does have a method, and may eventually share it with us, although
it is not a total solution. I believe it is built principally on spectral
methods for artifact detection used in EEGLAB.

2. If you haven't, see ADJUST's metrics and logic, something similar to
which
would be good to include. For example metrics such as SCADS.
As automated artifact detection methods are crucial, new toolboxes,
and comparisons between methods within these toolboxes, will
become more evident in the literature soon. See also
Dien's toolbox cleaning methods and output to think about
how to clean continuous data. There's also a growing ieee/signal analysis
literature on quasi-automatic ICA based artifact rejection.

3. Some groups have been working on automated methods
that mimic cleaning by real users.

4. A method that passes through the data at various
"fake" epoch lengths would be of interest.

5. a method that catches large, medium, small artifacts would be nice.

6. a method that allows catching/marking  of
eye blinks, eye movements, muscle tension, sweating, neck tension, brow
tension, jaw tension, etc.. would be nice.
A method that automatically builds up and uses subject-specific scalp maps
or ICs,
or relies on a library of such relatively spatially specific
artifact-related scalp maps, would be ideal.
Regarding muscle artifact detection, a recent paper from Davidson's group by
Schackmann comes to mind.

7. catching bad time samples of various lengths, at some or few or all/most
channels, would be nice.

8. Calibrating for global differences in signal across participant data sets
and/or recording systems would be good.

9. Being able to automatically identify different kinds of bad channels
automatically would be great.
e.g., noisy (i.e., 60hz) channels, flat channels, abnormal value channels,
channels with high variance.
Assumedly this would occur across the sweeps at various epoch sizes
mentioned above.
It's possible that channel by channel based interpolation would be good too,
automatically, from healthy channels and time periods.
Assuming an automated artifact detection method




Tarik Bel-Bahar
Swartz Center for Computational Neuroscience
Institute for Neural Computation, University of California, San Diego
San Diego SuperComputer Center, B-1, RM B191E
9500 Gilman Drive, MC: 0559 / La Jolla, CA 92093-0559




On Thu, Jun 9, 2011 at 7:53 AM, James Desjardins <jdesjardins at brocku.ca>wrote:

> Dear list members,
>
> For my purposes it is beneficial to reject artifacts in the continuous
> data rather than following segmentation. I find the artifact detection
> tools in EEGLAB extremely helpful when working with segmented data.
>
> I am considering working on a plugin that windows continuous data,
> takes advantage of some of the currently available artifact detection
> tools for segmented data, then translates the windowed rejection
> information back to time intervals in the continuous data.
>
> Are there already methods for doing this... or is this something that
> someone else is already working on?
>
>
> James Desjardins
> Technician, MA Student
> Department of Psychology, Behavioural Neuroscience
> Cognitive and Affective Neuroscience Lab
> Brock University
> 500 Glenridge Ave.
> St. Catharines, ON, Canada
> L2S 3A1
> 905-688-5550 x4676
>
>
>
>
>
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to
> eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to
> eeglablist-request at sccn.ucsd.edu
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://sccn.ucsd.edu/pipermail/eeglablist/attachments/20110610/ab923ca7/attachment.html>