ENH: basic support for eye tracker data

[dominikwelke]

Describe the new feature or enhancement

hi all,

for the 2022 sprint I suggested to add basic support for eye tracking data to MNE.
these data could be useful for several purposes, from improving data cleaning/IC labelling to all types of new analyses s.a. saccade related potentials.

@drammock suggested to post this issue ahead of time to discuss the topic.
I'd be happy to hear your feedback!

Describe your proposed implementation

there are several steps that I think would be necessary for a minimal implementation.

• deciding how to store these data it in the MNE ecosystem
  e.g. simply as a misc channel type along the other sensor data?
• handling of different sampling frequencies of the devices and missing data
  (as recording periods might only imperfectly overlap)
• basic I/O tools
  loading raw eye track data and/or annotiations (blinks, saccades) from different file formats (at least from text/csv files)
• alignment methods
  e.g. based on common annotations/trigger

a fuller (but still basic) version could also do (parts of) the following:

• allow to identify events from the raw eye track
  most importantly blinks, and saccades (e.g. via the Engbert&Mergenthaler algorithm)
• basic visualization tools

..and then of course there would be a host of specific tools the gaze information might be used for.
this might be beyond the scope of this issue, though.

Describe possible alternatives

..not even opening this box ;)

Additional comments

• a comment related to my other suggestion DOC: new tutorial for using automated preprocessing pipelines / switching back and forth between EEGLAB and MNE-Python #10750 : there is e.g. Olaf Dimigen's EEGLAB toolbox EYE-EEG that can do most of the above described. so if MNE could store eye-track data in a consistent way, it might also make sense to update the import functions
• I know there are several python toolboxes that act as parsers for different eye tracker data formats. it might make sense to check these out

[larsoner]

deciding how to store these data it in the MNE ecosystem e.g. simply as a misc channel type along the other sensor data?

+1 for a new channel type in the FIF standard

handling of different sampling frequencies of the devices and missing data (as recording periods might only imperfectly overlap) ... alignment methods e.g. based on common annotations/trigger

Already handled by https://mne.tools/stable/generated/mne.preprocessing.realign_raw.html

basic I/O tools loading raw eye track data and/or annotiations (blinks, saccades) from different file formats (at least from text/csv files)

I would start maybe by adapting the I/O from https://github.com/pyeparse/pyeparse, which isn't really maintained anymore (and should be deprecated in favor of proper MNE-Python support)

[larsoner]

I think we can close this as we have some basic support now, feel free to open new issues for additional functionality!