BrainVision event parser doesn't handle my data correctly

[hoechenberger]

The event parser for BrainVision files generates seemingly arbitrary event codes for my data. This is a problem that was first mentioned in #6267 and supposedly fixed. However, the fix introduced special handling depending on the detected event type:

https://github.com/mne-tools/mne-python/blob/70789c466133eaba410e1fc254c4b88723e30b77/mne/io/brainvision/brainvision.py#L825-L829

In the vmrk files generated in our lab using PyCorder, all stimulus and response events are simply labeled as Event (i.e., not Stimulus or Response, respectively). Therefore, they get assigned IDs 10001+, and the IDs we assigned to them during recording are basically ignored, yielding the issue(s) brought up in #6267.

When I change the line

_BV_EVENT_IO_OFFSETS = {'Stimulus/S': 0, 'Response/R': 1000, 'Optic/O': 2000}

to include Event markers:

_BV_EVENT_IO_OFFSETS = {'Stimulus/S': 0, 'Event/': 0, 'Response/R': 1000, 'Optic/O': 2000}

all works as expected and I can do

data = mne.io.read_raw_brainvision('data.vhdr')
events, _ = mne.events_from_annotations(data)
event_id = {'fixation': 100,
            'stim_a':   110,
            'stim_b':   120,
            'stim_c':   160,
            'response': 230}
epochs = mne.Epochs(data, events=events, event_id=event_id)

as would be expected.

This was tested with master.

x-ref https://github.com/cbrnr/mnelab/pull/50 cc @ArndalAndersen

[larsoner]

Seems reasonable to me. Do you happen to know if Event is basically the same thing as Stimulus, or are they different? Maybe just from different versions of PyCorder?

[hoechenberger]

Seems reasonable to me. Do you happen to know if Event is basically the same thing as Stimulus, or are they different? Maybe just from different versions of PyCorder?

I'm not sure, maybe it's just a configuration setting. The data were recorded two years ago and I currently don't have a working EEG setup to run further tests. But I could try to get a copy of PyCorder working again.

I would, intuitively, assume that Event is just a generic term for pretty much all kinds of stimulus and response events.

[hoechenberger]

@larsoner Maybe your question was actually whether the offset for Events should be zero or something else? The answer is, I believe it should always be zero in that case. :)

[agramfort]

I am fine with the suggested 1 line change.

you'll need to update a test though

thx

[hoechenberger]

@agramfort

you'll need to update a test though

I'd add a new sample vmrk file to mne/io/brainvision/tests/data and add a very small test that just tests whether Event markers are being read correctly, nothing else, would that suffice?

[larsoner]

Maybe just modify an existing vhdr file to have Event/ rather than Stimulus/, that way you don't even need to add a new file (or even a new test, since it should show up the same way)

[hoechenberger]

Maybe just modify an existing vhdr file to have Event/ rather than Stimulus/

You mean, replacing some existing Stimulus/S markers with Event/?

[agramfort]

You mean, replacing some existing Stimulus/S markers with Event/?

yes