[FieldTrip] EEG dataset with multiple electrode caps

[Schoffelen, J.M. (Jan Mathijs)]

Dear Soren,

Tricky business, I’d say. Let’s assume that you have  a more or less balanced number of subjects that were recorded with each of the caps. This at least would allow you to study the spatio(spectro)temporal response (and derived measures, such as the correlation with a behavioral variable) per subset, as well as to investigate the effect of ‘group membership’ on the result.

I think that each of the options are viable, but I would go for option 2). Your electrode caps seem to be sufficiently spatially dense to allow for spatial interpolation, for instance using ft_channelrepair (treating the target electrodes as the ‘missing’ channels). I would treat both groups of subjects with a spatial interpolation, rather than interpolating one of the groups onto the other one. Also, I think I’d try to avoid a source reconstruction step if spatial interpolation could do the trick.

It is hard to make a definitive statement about the ‘accuracy’ of each of the methods in general, but I’d think of both of them as a mapping procedure that projects the data into a ‘common space’, which subsequently allows for treating the individual data points’ spatial location to be comparable across subjects, which is a prerequisite for the spatial clustering.

Best wishes,
Jan-Mathijs


> On 8 Jul 2019, at 22:36, Soren Emmanuel Wainio-Theberge <swain083 at uottawa.ca> wrote:
> 
> Hello all, 
> 
> My name is Soren Wainio-Theberge, and I'm working in the Mind, Brain Imaging and Neuroethics unit in Ottawa, Canada. I have a dataset where we switched electrode caps partway through the experiment in order to obtain fNIRS data on some subjects. There is some limited overlap between the caps (a cluster of four electrodes in the parietal region and two in the frontal), but for the most part the electrode locations are different. I'm now looking to use cluster-based permutation testing with the whole dataset, testing a correlation with a psychological variable. Is this possible? If so, what would be the most sound approach? I can see three options at the moment:
> 
> 1) Convert all subjects with the fNIRS cap to the original cap by interpolating those electrodes. 
> 2) For each cap, interpolate the electrodes of the other cap to get a common space with more electrodes. 
> 3) Do the analysis in source space to avoid the electrode cap issue altogether (though we only have 64 electrodes for each cap and no individual anatomy, so source estimation won't be very accurate). 
> 
> Which is the best option from the perspective of a) the cluster test and multiple comparisons (ie does option 2 cause more issues for the cluster testing than option 1), and b) the accuracy of the method in general ( interpolation vs source space)? Or is there another possibility which I haven't thought of?
> 
> Thanks very much, 
> Soren
> _______________________________________________
> fieldtrip mailing list
> https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
> https://doi.org/10.1371/journal.pcbi.1002202


_______________________________________________
fieldtrip mailing list
https://mailman.science.ru.nl/mailman/listinfo/fieldtrip
https://doi.org/10.1371/journal.pcbi.1002202