MRG, ENH: BrainVision impedance parsing

[sappelhoff]

takes over #7771
closes #7771

notes for posterity

• I looked at this vhdr file to adjust some of our own test files: https://gin.g-node.org/JiriVanek/DataConversionToNIX/src/master/BrainVision%20format/RAW/P3Numbers_20150618_f_10_001.vhdr

[agramfort]

I am wondering if we should think about adding support for this in the fif standard.

@larsoner @mkajola jnenonen any thought on adding EEG impedance values to fiff format?

[sappelhoff]

The tests are failing because object_diff is comparing np.nan as False:

In [4]: np.nan == np.nan                                                        
Out[4]: False

so actually, everything is passing. How should I fix this?

EDIT: Okay, I think I found why --> object_diff did so far not deal with non-array np.nan floats. Should be fixed now.

[agramfort]

the self.impedances attribute should be added to the attributes documentation of the Raw class.

see https://mne.tools/dev/generated/mne.io.Raw.html#mne.io.Raw

if should be None if not present. I would like to have a special case of a Raw where code work with a RawBranvision instance but breaks with RawEDF etc.

the common test for raw object should also be amended.

how do you see the pb of now saving the attributes on the file with raw.save?

[sappelhoff]

these points go back to the discussion we had some weeks ago.

This is what @larsoner said about this in his comment

BrainVision raw files will be the only files to have this attribute?

I think this is okay. It will also be lost on FIF round-trip but that's also okay. If there are other readers that support impedances we can think about a more global way of storing them or something. As long as we pick a sensible attribute here we can always extend it to other readers later via a mixin or something.

---

If raw.impedances becomes a publicly documented attribute of Raw, then we also HAVE to implement proper saving of this attribute, right?

So the options I see:

1. make it an undocumented attribute of BrainVision files only, until we get more file format readers to parse impedances
2. make it a fully documented attribute of Raw, with possible FIF roundtrip

With option 2., I am a bit uncomfortable, because I don't know how many file formats even support saving the impedance values in their file formats?

I have only ever seen this for BrainVision files. Do you know of more @hoechenberger @cbrnr ?

[agramfort]

does eeglab set files have a documented way of storing this info?

…

[sappelhoff]

does eeglab set files have a documented way of storing this info?

I did a quick google search, and not alot shows up. But I am no expert on .set

EDF files also do NOT seem to have impedances: https://www.edfplus.info/specs/edfplus.html

[hoechenberger]

I have only ever seen this for BrainVision files. Do you know of more @hoechenberger @cbrnr ?

I've only worked with BrainVision and BioSemi data, and I only remember seeing this for BV!

[cbrnr]

I haven't seen that either. If you want to make it a documented attribute, then the default could be None (meaning for all formats except BrainVision). Even BrainVision files normally don't have this kind of information, so it should really be an optional piece of information.

[hoechenberger]

Even BrainVision files normally don't have this kind of information

Oh really? Interesting, and good to know! The ones produced in our lab always contained this info. Do you know under which circumstances this info will / will not be produced?

[cbrnr]

Maybe I'm wrong, but this should only work for amps that measure impedances (i.e. active electrodes) - so you won't find this info for passive amps. Also, none of our example files in the testing data repo has impedance info.

[hoechenberger]

Maybe I'm wrong, but this should only work for amps that measure impedances (i.e. active electrodes)

Ah that could be right, I've only used active electrodes in the past years!

[sappelhoff]

whether or not impedances are saved with the BrainVision files really depends on a lot of things.

• the amplifier/electrodes being used
• the recording software and settings being used (BrainVision Recorder works, but also not in all settings/impedance-checks)
• whether or not you have a USB connection to the "actiCap control box"(interface between amps and electrodes) vs only the fibre optics cables

For me active vs passive was not a deciding factor --> I can record impedances from both activa AND passive electrodes (see e.g., the HEOG+ and HEOG- impedances I added in this PR --> that's real data from passive electrodes)

---

anyhow, it seems like reading impedance values from EEG file formats may prove to be difficult.

possible	not possible
BrainVision (sometimes)	EDF
	BDF
	.set

What about FIF?

[larsoner]

Not that I know of

[agramfort]

can you at least update https://github.com/mne-tools/mne-python/blob/master/mne/io/brainvision/brainvision.py#L52 to document attributes?

then let's reconsider making impedances first class citizens in mne when more formats support this...

[larsoner]

FYI we don't document the individual classes so this will only show up if you look at the doc of this class __init__ in a terminal or IDE (which is probably slim to never). But I guess we can keep it

[larsoner]

Thanks @sappelhoff

[sappelhoff]

seems like the GDF format supports impedance values: https://arxiv.org/pdf/cs/0608052v6,

but I have never used that format and have not seen it "in the wild" too much.