Fwd: Fieldtrip - MNE followup

[jan-mathijs schoffelen]

Dear Geoff,

> - When looking through the ft_write_data file, I don't have 'cft_ds'
> as an
> option to the dataformat parameter.  is this an older option which was
> lost?

You're absolutely right. The option to write ctf-data is not included
anymore in ft_write_data. We should update the documentation
accordingly. I just browsed the fieldtrip-wiki and found this link,
without properly checking whether it was still up to date. Sorry about
that.

> - The documentation is a little unclear about what sort of data is
> to be
> used as input to the function.  I've tried using the output of
> timelockanalysis, as well as the 'avg' field of the same structure.
> For
> the former, I get an error that the function is not defined "for input
> arguments of type 'struct'", and for the latter that it's not
> defined "for
> input arguments of type 'char'".  I think the latter is getting
> confused
> about the filename input, then.  When I give it a header object
> taken from
> the original input .ds file with ft_read_header and modified to fit
> the
> changes in analysis, I again get the 'struct' error.  I've been
> playing
> around with all the options and looking into the script, but I can't
> get
> anything to work.  Do you happen to have a working script handy?

No, I don't have a script lying on the shelf, because I don't
generally go back to a raw file format after having processed my data
in fieldtrip. Since the use of ft_write_data does not seem to support
what you want to do anyway, I will not look in detail at the function,
but I guess it needs a filename (as a string), a data matrix (as a
matrix), and some additional inputs, such as requested fileformat etc.

> Also, I noticed your email from a few days back, called "warning for
> ctf-users".  If I only performed the normal analyses (i.e.
> timelockanalysis, megplanar, combineplanar, timelockbaseline,
> timelockgrandaverage), do you think I should worry about redoing the
> analysis with the new scripts?  I think I'm safe, but thought it
> might be
> a good idea to check with an expert!

With respect to my warning issued earlier: you are safe. As long as
you did not explicitly use the function ft_denoise_synthetic yourself,
there's no reason to worry.
Yet, I have another warning for you: given your analysis steps it is
not possible to write the data back to CTF-dataformat in a consistent
way, since you transformed the data to planar gradient representation
(which results in (and requires) a totally different description of
the sensor array; so if you plan to use a different package for you
minimum norm estimate there will be for sure a mismatch between the
channel data and the leadfields), and you have used combineplanar. The
latter is a nonlinear operation on your data, which is definitely not
allowed if you would like to compute leadfields using the sensor
information as it is represented in the original .res4 header file.

As a side remark I quickly looked into writeCTFds, and it seems in
theory possible to use this function to write the processed data to a
CTF-dataset (as long as one stays in the original axial gradient
representation). Yet, I wonder whether it will be worth the time
investment to get it smoothly implemented (because it is not totally
non-trivial). I'd rather opt for looking into the various other
suggestions issued in this thread (by Nathan and Vladimir for
example), and invest some time there, or just give fieldtrip's
minimumnormestimate a try. Keep always in mind that the planar
transformation as such requires a change in the description of the
sensor array (which is done by fieldtrip), and that combining planar
gradient data before going to source space is not good.

Cheers,

Jan-Mathijs

----------------------------------
The aim of this list is to facilitate the discussion between users of the FieldTrip  toolbox, to share experiences and to discuss new ideas for MEG and EEG analysis. See also http://listserv.surfnet.nl/archives/fieldtrip.html and http://www.ru.nl/neuroimaging/fieldtrip.