EEG processing pipeline with autoreject

💬 Support & Discussions
, , , ,
3

Hi Eduardo,

We recently created a tutorial in autoreject with the help of Alex Rockhill, just to address these kinds of questions:

https://872-59485140-gh.circle-artifacts.com/0/html/auto_examples/plot_autoreject_workflow.html#sphx-glr-auto-examples-plot-autoreject-workflow-py

Specific answers:

  1. Yes, autoreject after ICA is the correct order. You can also apply it before, and use the bad epochs detected to improve ICA decomposition. Local autoreject before ICA seems to work better than global autoreject. If you’re lazy, you could also just use a high threshold like 500 uV pre-ICA. I haven’t investigated in detail the impact of these different choices, so make sure to inspect your data.

  2. See 1.

  3. I’m not exactly sure what kind of features you’re talking about. If you intend to do some machine learning, I would normalize the features, yes.

Best,

Mainak

2 Likes