[FieldTrip] Using ICA to filter data

frank@greenant.net fieldtrip at greenant.net
Sat Jun 18 08:56:57 CEST 2011


My apologies, I think I may have been a bit unclear about the aims of collecting the data.
We're aiming to measure saccadic onset from the signal on a per-trial basis.
I was hoping to isolate the MRI noise "component" and subtract this from the EOG signal component
to give a cleaner EOG signal.

Can the ICA module be applied in the temporal domain in this fashion?
There are 40 single session trials of 6 seconds each, I would have thought that'd be enough data to
give it a try.  Also open to any suggestions as to other adaptive filter techniques....

Thanks

On 17/06/2011, at 8:37 AM, Rodolphe Nenert wrote:

> Im still afraid that the power of this analysis will be very low.
> Maybe you can try an ICA on your fMRI data and try to correlate each component with your EOG timecourse.
> 
> Rodolphe N.
> 
> On Thu, Jun 16, 2011 at 1:16 PM, frank at greenant.net <fieldtrip at greenant.net> wrote:
> It's a bit of a unique experiment, we're trying to use an ECG machine
> to acquire EOG, so it's only a single output channel.
> 
> i guess what I need is a temporal ICA rather than a spatial one...
> 
> On 17/06/2011, at 2:48 AM, Rodolphe Nenert wrote:
> 
>> To summarize, the ICA will decompose your signal into as many components as Electrodes.
>> Therefore, trying to decompose only one source is useless.
>> Did you use a full net of electrodes into your MRI machine or only EOG electrodes?
>> 
>> Hope this helps, 
>> 
>> Rodolphe N.
>> 
>> On Thu, Jun 16, 2011 at 11:42 AM, frank at greenant.net <fieldtrip at greenant.net> wrote:
>> I am trying to use fieldtrip to filter EOG data obtained in an MRI.
>> I want to be able to spot the saccades in the samples and ideally measure their onset at the
>> end of each trial.
>> 
>> As you may guess, it's quite noisy and it's broad spectrum noise, despite pre-filtering.
>> 
>> Ideally, I would like to isolate the component that corresponds to the MRI interference
>> and then filter this out.
>> 
>> I have managed to import the data and can run ft_componentanalysis
>> but it fails with:
>> 
>> runica() - data size (1,30720) too small
>> 
>> My data is single channel, 40 epochs, each of 6 seconds (time locked to stim onset but not saccade onset)
>> Is there a different method I should be using?
>> 
>> I have posted some sample data and the current script (which reads in the
>> data and runs preprocessing) to the following urls:
>> 
>> http://greenant.net/temp/1_2_MRI_2011-04-29%2016:19:33.mat
>> http://greenant.net/temp/EOG_analysis.m
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
>> 
>> _______________________________________________
>> fieldtrip mailing list
>> fieldtrip at donders.ru.nl
>> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> 
> 
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> 
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20110618/95750f9c/attachment-0002.html>


More information about the fieldtrip mailing list