Mne-python: Creating stimulus channel from scratch / adding events to the raw object

[ryscet]

I have data from experiment in two different files, first and .edf file with EEG data and second a csv file with events. I would like to create a mne.Raw (mne.io.read_raw_edf()) object and afterwards add a stimulus channel to it.

From the csv file I created an array of events which I would like to add the the mne.Raw using raw.add_events(). This however requires a stimulus channel which is not there, so it needs to be created. I couldn't find an example or documentation describing how to create a stimulus channel from scratch or more generally on adding events from another file. Is events channel an array of 0's plus a value coding for event on all event indexes?

I understand I could create epochs data using the raw object created from edf file and events from csv file but for convenience I would like to store them in a single mne object, is there such possibility?

[jaeilepp]

You can use add_channels

So the flow would be something like this:

info = mne.create_info(['STI'], raw.info['sfreq'], ['stim'])
stim_raw = mne.io.RawArray(stim_data, info)
raw.add_channels([stim_raw], force_update_info=True)

[ryscet]

Okay, thanks! How should stim_data look like?

[jaeilepp]

It should be 2d array. So something like

stim_data = np.zeros((1, len(raw.times))

[ryscet]

Great, that worked. I didn't know what was the format of the stim_data, now it's clear. Perhaps this should go into documentation, I could make an example of how to add events from a separate file and a submit a pull request.

[larsoner]

That would be helpful @ryscet

[piotrdzwiniel]

It would be great to include in the documentation information about the required shape of raw objects send to add_channels function in add_list list parameter. When adding only one channel it is not clear to represent this one-dimensional channel as two-dimensional matrix n_channels, n_samples as for example (1, n_samples). It's not intuitive.