load discontinuous Neuralynx data as epochs

[ludovicbellier]

Hi there,

As discussed on issue #1128, here's a way to automatically load discontinuous Neuralynx data as epochs (and to avoid getting repeated, artifactual samples in our data due to the pause/stop built-in feature).

Best,
Ludovic

[robertoostenveld]

you cannot suddenly add a field to the output of ft_read_header and start using that here. That breaks the API, which is shared with EEGLAB, SPM and other software. See http://www.fieldtriptoolbox.org/development/module/

[ludovicbellier]

Ok good to know. Thanks to your suggestion I created ft_trialfun_neuralynx.m, and besides the fact that I now have to add a few lines before my call to ft_preprocessing, this is indeed way cleaner!

[robertoostenveld]

this is fine if all trials have the same length.

[robertoostenveld]

please note that hdr.nSamples*hdr.nTrials should correspond to the total number of samples in the file.

[robertoostenveld]

you cannot add fields to the hdr structure.

[robertoostenveld]

One solution would be an implementation similar to https://github.com/fieldtrip/fieldtrip/blob/master/trialfun/ft_trialfun_brainvision_segmented.m
which is also a file format that can have segments of different length.

But I believe best would be to use the existing ft_read_event output and add the segments as "events" (with the appropriate event.duration). That should allow you to do something similar to

cfg = [];
cfg.dataset = xxx
cfg.trialfun = 'ft_trialfun_general';
cfg.trialdef.eventtype = 'segment';
cfg = ft_definetrial(cfg)  % this adds cfg.trl, only for the segments

data = ft_preprocessing(cfg); % this reads the segments

This is also how "trial" events are being read from some files, such as the CTF format. E.g. look at Subject01.ds, which has 3-second (900 sample) segments.

[ludovicbellier]

Hey Robert,

Thanks for your comments!

I first explored your second (best) suggestion, but it bugged at line 1732 of ft_read_event while trying to load an event (.nev) file that in my case doesn't exist (my cfg.dataset is a directory which only contains .ncs files).

So I switched to your first suggestion and created ft_trialfun_neuralynx.m, basically with the code I previously inserted in fileio/private/read_neuralynx_ds.m, and it works great. However, we should remove the two previously modified files (ft_preprocessing.m and fileio/private/read_neuralynx_ds.m) from this pull request; could you please take care of this, as I am not sure how to do it without messing around with the master branch?

Please let me know if you think we should make any other changes.
Best,
Ludovic

[ludovicbellier]

Ok good to know. Thanks to your suggestion I created ft_trialfun_neuralynx.m, and besides the fact that I now have to add a few lines before my call to ft_preprocessing, this is indeed way cleaner!