[FieldTrip] follow-up on grad fields with NaNs after ICA

[Vitória Magalhães Piai]

Dear Vitória,

The short answer to your question would be:

If in subsequent steps you do planar gradient transformation or data 
visualization it's safe to replace the grad. If in subsequent steps you 
want to do sourceanalysis, it's not safe to replace the grad.

Reason: essential information that is needed for a correct forward 
computation (i.e. construction of the leadfields) is removed by 
replacing the grad. This information is represented in grad.tra. The 
explanation would be a bit technical, and it has been our intention for 
a while to document this on the FT-wiki as a FAQ, but we didn't get to 
it yet. For visualization and planar gradient transformation, only the 
grad.chanpos is needed, and this is correctly represented in an 
'earlier-stage' grad-structure.

Note that we have made some changes just this past week that actually 
addresses this NaN-issue. It will not work on backprojected data (i.e. 
data that has gone through ft_componentanalysis, followed by 
ft_rejectcomponent), but if you grab the most recent FT version and do 
ft_rejectcomponent, I believe the grad.chanpos will be reconstructed 
without NaNs. I haven't tried it myself, but rumour has it that this is 
the case.

Would you be interested in contributing to the documentation, let me know...

Best wishes,

Jan-Mathijs

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

On Oct 5, 2013, at 4:26 AM, Vitória Magalhães Piai wrote:

> Hi FT-ers,
>
> It took me a while to find some older posts that would answer my 
> question, and this thread almost did the whole job:
> http://mailman.science.ru.nl/pipermail/fieldtrip/2013-July/006779.html
> http://mailman.science.ru.nl/pipermail/fieldtrip/2013-July/006788.html
>
> Like Alex, I figured out that it was the ICA that left my grad.chanpos 
> with NaNs only, and like Haiteng, I thought just replacing the grad 
> field with a pre-ICA grad field would be one way to go.
> But just like Alex I was wondering whether this is the right way to 
> go: "however, it feels somewhat dangerous to perform such "dirty" 
> fixes, since that grad info is probably removed for a certain reason. 
> We are oblivious to its impact subsequent steps like planar transform 
> and source reconstruction."
> Talking to a knowledgeable colleague yesterday about this issue, he 
> also had his question marks. But I don't see any follow-up on Alex's 
> remark (I guess he further discussed this with colleagues while 
> drinking coffee at the DCCN kitchen...! :-)
> SS
> So I still don't know the answer and maybe some other people may want 
> to know it as well in the future.
>
> Any opinion on whether it's safe to use pre-ICA grad information for 
> subsequent planar transformation and source analyses?
> Thanks a lot, Vitória
>
> -- 
> ** Please consider the environment - do you really need to print? **
>
> _______________________________________________
> fieldtrip mailing list
> fieldtrip at donders.ru.nl
> http://mailman.science.ru.nl/mailman/listinfo/fieldtrip

Jan-Mathijs Schoffelen, MD PhD

Donders Institute for Brain, Cognition and Behaviour,
Centre for Cognitive Neuroimaging,
Radboud University Nijmegen, The Netherlands

Max Planck Institute for Psycholinguistics,
Nijmegen, The Netherlands

J.Schoffelen at donders.ru.nl <mailto:J.Schoffelen at donders.ru.nl>
Telephone: +31-24-3614793

http://www.hettaligebrein.nl



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.science.ru.nl/pipermail/fieldtrip/attachments/20131005/2a0e04f8/attachment-0002.html>