[Eeglablist] Seeking your advice on EEG preprocessing!

[Zigmunds Freibergs]

Dear EEGLAB community,



I would appreciate your feedback on the EEG preprocessing pipeline I have
created based on the literature analysis (I am a first-year experimental
psychology PhD student with only 2 years of mostly EEG data-gathering
experience). My main goal is to ensure that the pipeline is
methodologically sound, especially regarding ICA, artifact cleaning, and
standardization for larger datasets. And I am working with Biosemi, g.tec,
and ANT data files.



*Pipeline:*

   1. Uploading channel location file
   2. Initial average re-referencing
   3. (Optional) downsampling
   4. Before filtering - dataset duplication for ICA training
   5. Steps on original files

5.1. Cleanline – to remove line noise

5.2. Band-pass filter: 0.1-30 Hz (or 40Hz? (For ERP`s)

5.3. Automated bad segment/channel handling (clean_rawdata / ASR) – using
Makotos suggested approach, to unthick remove bad channels, and use a max
acceptable 0.5 s window, standard deviation 30 (for me 20 seems still more
aggressive in removing data segments)

5.4. interpolation of channels

5.5. rereferencing to average.

   6. Steps on duplicated files

6.1. Band-pass filter: 1 (not 0.1) –30 Hz

6.2. Run ICA

   7. Further steps performed back on original files, after step 5.5.:

7.2. Transfer ICA weights from ICA-training copy

7.3. IC classification and rejection

7.4. Epoching, baseline correction etc.

My questions are:

1.      Is the overall logic and sequence of this pipeline correct?

2.      What preprocessing steps should strictly precede ICA, and which
should not (and why)?

3.      How do you approach artifact cleaning in a standardized yet
data-preserving way?

4.      Do you recommend clean_rawdata / ASR, alternative automated
methods, or manual inspection - and why? (And which methods?)

5.      At which stages do your own pipelines differ from the one outlined
above, and what is the rationale?

6.      Is it ok to do initial re-referencing to the average? Why yes or no?

7. At which stage of preprocessing is it best to perform epoching, and what
is the rationale? Should it be done before or after ICA and artifact
cleaning, and how does the choice affect ICA quality and ERP analysis?



Thank you very much in advance!



Best regards,

Zigmunds

<https://urldefense.com/v3/__https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail__;!!Mih3wA!ElVgq-unr0kgp_I1VuXGLJXTrE48Ca0zjSy8KVE0JoHhIf-kFR5sjYU_e3ms4sS1cB6xUCqH8mnug8YSJeiPwqkbWwD9CfWm$ >
Virus-free.www.avast.com
<https://urldefense.com/v3/__https://www.avast.com/sig-email?utm_medium=email&utm_source=link&utm_campaign=sig-email&utm_content=webmail__;!!Mih3wA!ElVgq-unr0kgp_I1VuXGLJXTrE48Ca0zjSy8KVE0JoHhIf-kFR5sjYU_e3ms4sS1cB6xUCqH8mnug8YSJeiPwqkbWwD9CfWm$ >
<#DAB4FAD8-2DD7-40BB-A1B8-4E2AA1F9FDF2>