`run_unfold` throws an error if the data does not only include eeg (but for example also eye-tracking data).

[jschepers]

I tried using UnfoldBIDS with our co-registered EEG + ET data and run_unfold throws the following error:

I looked into the stacktrace and the problem is that the internal function raw_to_data calls the MNE get_data function with "uV" as units argument.

UnfoldBIDS.jl/ext/MNEext.jl

Lines 30 to 37 in 3756e66

	raw_to_data(raw; channels::AbstractVector{<:Union{String,Integer}}=[])
	Function to get data from MNE raw object. Can choose specific channels; default loads all channels.
	"""
	function raw_to_data(raw; channels::AbstractVector{<:Union{String,Integer}} = ["all"])
	return pyconvert(Array, raw.get_data(picks = pylist(channels), units = "uV"))
	end

However this does not work if there are not only eeg data channels.

Possible fixes could be:

1. Use pick on the Raw object beforehand to only choose eeg channels.
2. Use picks within the get_data function and set it by default to "eeg".
3. Inform the user that they have to remove other channels from their data (but I don't like this option).

Cheers,
Judith

P.S. I don't have time to do a PR right now, but I could do it in two weeks from now. But I wanted to hear your opinion about the issue first anyway :)

[ReneSkukies]

The raw_to_data function is supplied with the channels keyword and kwargs from run_unfold. So if you run_unfold(data_df, bfdict; channels="eeg") it should work as intended.

Otherwise, you can also supply your entirely own data extraction function to run_unfold via the extract_data keyword =)
Link to relevant docs

[jschepers]

@ReneSkukies Thanks for the quick response and the workaround. However, I think it would be nice if a user got at least a more informative error message in that case such that they know that they have to set channels="eeg" (if you don't want to set EEG as default).

[jschepers]

The raw_to_data function is supplied with the channels keyword and kwargs from run_unfold. So if you run_unfold(data_df, bfdict; channels="eeg") it should work as intended.

Otherwise, you can also supply your entirely own data extraction function to run_unfold via the extract_data keyword =) Link to relevant docs

Setting channels="eeg" did not work and gave the following error

becuase raw_to_data expects a vector. channels=["eeg"] works.

[ReneSkukies]

Yeah sorry, forgot the vector, but that is by design, because it is in line with MNE (picks also expects a vector afaik).

I don't think it's clear to me how setting channels = ["eeg"] as the default would be beneficial as that would then throw an error in for example MEG datasets.
But I see the incompatibilities there and can provide an assert check in the function with a more meaningful error and additionally the ability to change the units.

[ReneSkukies]

• implement assert to check for eeg channels or lack thereof
• make units a kwarg

[jschepers]

You mean the get_data function in MNE? I think picks can either be a string or a array_like: https://mne.tools/stable/generated/mne.io.Raw.html#mne.io.Raw.get_data

I see your point that you want to keep the code as general as possible to also be able to process other data types than eeg. However, then I also think that "uV" is not an appropriate default because it assumes that you process eeg data (or at least data that is measured in uV).

Thank you!