Logical problem with CTF 3rd order gradiometers correction

[Robert Oostenveld]

Hi Michael

On 31 Jul 2008, at 18:11, Michael Wibral wrote:
> Dear listusers,
>  I have a logical problem with denoise-synthetic.
>
> here's  what I want to do:
>
> 1. use definetrial to make cfg.trl (works fine)
> 2. use preprocessing with the new cfg. headerfile='ctf_new' ,
> cfg.datafile='ctf_new' options to read the reference gradiometer
> information and the data into a fieldtrip structure (works fine)
> 3. use denoise_synthetic to denoise the data (works fine ?)

In the (sofar limited) experiences that I heard there are noticeable
differences mainly on the central/vertex sensors (which are also the
closest to the reference sensors).

> 4. un artefact detection on these data (-> this works fine and
> updates cfg.trl)

after artifact_xxx and rejectartifact you should read the data using
preprocessing. Or are you using rejectvisual? If so, then you don't
have to call preprocessing a second time, because the output of
rejectvisual is again your data with the bad trials removed.

> 5. run preprocessing a second time for filtering and actually
> applying the results of artefact corrcetion,... i.e. those things
> that I didn't want to do before denoise_synthetic.

After rejectartifact, you should be calling it here like preprocessing
(cfg), and not with a second input argument.

If you call it like preprocessing(cfg, data), then the cfg should not
contain cfg.dataset and cfg.trl but only filtering options, because
the data should not be read from file.

> BUT preprocessing complains that cfg shouldn't contain the cfg.trl
> field when no data are read - which is the case because I am
> invoking preprocessing a second time here on an existing fieldtrip
> structure.

I suspect that you are mixing up rejectartifact and preprocessing.
Calling preprocessing with a 2nd data input argument is only meant to
apply another set of filters on that data (i.e. if you forgot to do
it in the first run, or if you sequentially want to bandstopfilter
some line noise or CRT artifacts).

> Therefore all information from artefact rejection is lost.
>
> Somehow it seems that I can run denoise_synthetic only after
> artefact rejection, because I need a fieldtrip dataset to run it,
> but NO dataset for preprocessing to perform artefact correction,
> i.e. accept cfg.trl. This would be a pity.

I must say that I had not considered sofar the interaction between
denoising and artifact detectoin and removal. The artfifact_xxx
functions perform detection on data from disk(*), i.e. prior to
reading the data into memory. The historical reason for (*) is to
save on memory when a lot of filter padding is used and when there is
a lot of overlap between padded trials. The padded regions should be
checked for artifacts (esp jump artifacts which cause a lot of filter
rinning), but in the end we don't need them in memory. The denoising
only happens after the data is read into memory.

> Any suggestion how to solve this?

For the moment I suggest to use the following sequence

definetrial
preprocessing
denoise_synthetic
rejectvisual

and not use artifact_xxx combined with rejectartifact. We are
revising the automatic artifact detection functions at the moment,
since computers nowadays have much more RAM than when we started
implementing it 5 years ago, and reading the padded data into memory
is now much less of a concern. So in the near future, it will be
possible to do autodetection of artifacts also on data that is in
memory, which means that then you could also do the autodetection
after denoise_synthetic.

best regards
Robert

----------------------------------
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/fcdonders/fieldtrip.