pyAutomagic for automatic EEG preprocessing and quality assessment

[clairezurn]

What does this implement/fix?

There is a need for automatic and replicable preprocessing and evaluation of data quality in EEG to increase transparency in the field. A package for this purpose, Automagic, was developed in Matlab (https://github.com/methlabUZH/automagic). It provides the user an easy way to set preprocessing and quality evaluation parameters for an entire project, and automatically execute. The quality metrics used are all based on variance of some type in the data, and these metrics are used to give each recording a rating of "good", "ok", or "bad". All results, figures, and log files, are then saved.

Describe your solution

The suggested solution, pyAutomagic, is the movement of Automagic's functionality to python, specifically compatible with mne-python and mne-bids. The outcome would be a repository in mne-tools.

Additional information

More info on the original (MATLAB) package functionality can be found in the paper published (https://www.biorxiv.org/content/10.1101/460469v3.full).

[adam2392]

Let's also figure out a test dataset perhaps as well. Can we by any chance get one with expected output in MATLAB automagic to test against? Definitely we probably don't need exact matching, but would serve as a benchmark.

In addition, a solution should include (ideally) to be worked on w/ @saulmezac @deepsoni1996 :

1. some documentation for reference on what functions do, and meta-summary of the pipeline
2. some tutorials on how to run full pipeline, sub-parts of the pipeline, and then example analysis at the end
3. design change(s) summary compared to automagic.

@nickilanger any suggestions here?

[nickilanger]

Hi Andy, Yes. it’s a good idea to have a data set compared the two pipelines. I can provide the validation data set, which we used for the Neuroimage paper. Best, Nicolas Nicolas Langer Assistant Professor of Methods of Plasticity Research University of Zurich Andreasstrasse 15 (Office AND.4.56) 8050, Zurich, Switzerland n.langer@psychologie.uzh.ch phone: (+41) 44 635 34 14 http://www.psychology.uzh.ch/en/areas/nec/plafor.html

…

On Nov 15, 2019, at 6:36 PM, Adam Li ***@***.***> wrote: Let's also figure out a test dataset perhaps as well. Can we by any chance get one with expected output in MATLAB automagic to test against? Definitely we probably don't need exact matching, but would serve as a benchmark. In addition, a solution should include (ideally) to be worked on w/ @saulmezac <https://github.com/saulmezac> @deepsoni1996 <https://github.com/deepsoni1996> : some documentation for reference on what functions do, and meta-summary of the pipeline some tutorials on how to run full pipeline, sub-parts of the pipeline, and then example analysis at the end design change(s) summary compared to automagic. @nickilanger <https://github.com/nickilanger> any suggestions here? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AGGDUJE44JIS7QFYSI4KF6DQT3MYJA5CNFSM4JNWMNHKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEGFEWA#issuecomment-554455640>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGGDUJG7AZT7MZYZAEHRCJ3QT3MYJANCNFSM4JNWMNHA>.

[clairezurn]

We definitely need a BIDS compliant dataset to test on. The ones here (https://github.com/bids-standard/bids-examples) have empty data files, so are only helpful to the extent of making sure we are querying files correctly. @adam2392 or @nickilanger please let me know if you have suggestions on that. If nothing else already exists, I'll have to make this myself. I will begin the summary of changes compared to automagic.

[nickilanger]

The data we have are BIDS compliant. Best, Nicolas Nicolas Langer Assistant Professor of Methods of Plasticity Research University of Zurich Andreasstrasse 15 (Office AND.4.56) 8050, Zurich, Switzerland n.langer@psychologie.uzh.ch phone: (+41) 44 635 34 14 http://www.psychology.uzh.ch/en/areas/nec/plafor.html

…

On Nov 18, 2019, at 3:42 PM, clairezurn ***@***.***> wrote: We definitely need a BIDS compliant dataset to test on. The ones here (https://github.com/bids-standard/bids-examples <https://github.com/bids-standard/bids-examples>) have empty data files, so are only helpful to the extent of making sure we are querying files correctly. @adam2392 <https://github.com/adam2392> or @nickilanger <https://github.com/nickilanger> please let me know if you have suggestions on that. If nothing else already exists, I'll have to make this myself. I will begin the summary of changes compared to automagic. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2?email_source=notifications&email_token=AGGDUJD4QIZS3NJQZU7VEQ3QUKSWXA5CNFSM4JNWMNHKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEKU56Q#issuecomment-555044602>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AGGDUJCI3TSVKEBDNR5WW63QUKSWXANCNFSM4JNWMNHA>.

[adam2392]

Hi Nick,

Sounds great. Would you be able to put it in the Dropbox you shared with Jovo and I? This is publicly available, and we can include it inside the repo?

[clairezurn]

Any other feedback on phrasing this for submitting an issue to the real mne-python repository? I'd like to do so as soon as possible. I'll plan on mentioning in solution that it will be validated with the same dataset used for the Automagic repo and that documentation and tutorials will be included.

[adam2392]

This is good. Include all the comments and mention and link the datasets and original paper and you’re good.

Tag me so I can follow the convo.