How to import EEG-lab / .set EEG.events field for Unfold?

[behinger]

I just got asked how to best import EEG-lab style .set EEG.event fields.

[behinger]

Let's first download a dataset:

	using OpenScienceFramework 
        const OSF = OpenScienceFramework
        osf_proj = OSF.project(OSF.Client(), "wbz7x") 
        cp(readdir(osf_proj)[1], tempdir()*"/unfold_tempdata.zip") # copy / download
        run(`unzip  $(tempdir()*"/unfold_tempdata.zip") -d  $(tempdir()*"/unfold_tempdata_unzipped/")`) # unzip

Now that we have an eeglab-file, we can extract the events:

	using MAT
	using DataFrames
        EEG = MAT.matread("/tmp/unfold_tempdata_unzipped/face_saccades_opendata_fig10.set")
        evts = DataFrame([k=>dropdims(v,dims=1) for (k,v) in EEG["EEG"]["event"]])

Why so complicated with dropdims? Because MATLAB automatically saves vectors as 1xN matrices...

Tipp: It might be necessary to convert the columns to proper types (by default Any is used. This can be achieved by adapting some code like this:

	function converter(v)
		v2 = try 
					Float64.(v)
			 catch
					string.(v)
			 end
		return dropdims(v2,dims=1) 
	end
	evts2 = DataFrame([k=>converter(v) for (k,v) in EEG["EEG"]["event"]])
end

If this code helps you, please comment! Then others know they are on the right track :)

[EmmanuelleKris]

Thank you for this very helpful example.

It works without issue with the .set files from your tuto but with our data, we get :

ERROR: KeyError: key "EEG" not found

On the other hand, it works if we save the EEG structure from EEGlab in .mat format ...

[behinger]

Great!
Mhhh - maybe your dataset is saved as set / fdt?

Ah- oh! I forgot to include the line where I actually load the EEG file 🙈

EEG = MAT.matread("/tmp/unfold_tempdata_unzipped/face_saccades_opendata_fig10.set")

did you use the same?

[EmmanuelleKris]

Yes, same line.

[EmmanuelleKris]

And to load EEG.data, I can't do it from .mat file with this method :

data= DataFrame([k=>dropdims(v,dims=1) for (k,v) in EEG["EEG"]["data"]])
ERROR: BoundsError: attempt to access Float32 at index [2]

I can do it with PyMNE, as you show in the workshop, from .set file.

But this whole operation doesn't seem optimal...

[behinger]

yes, that line of code would not work for the data.

As you say, I would probably just get the data via PyMNE, but the events via the mat file.

or just use EEG["EEG"]["data"], that should already give you the data in the correct format :)