[Mne_analysis] EEG pipeline with MNE

[Davide Aloi]

Hi there,
I am analyzing some data from an experiment with children with dyslexia and
a control group. The experiment was done using a 128 electrodes EEG.
Basically, children were presented some words which could be wrong spelled
or correct and were asked to decide whether these words were correct or
not. The words were 120, 60 correct and 60 incorrect. These words were also
divided into different categories (included/not included in a previously
done training).

I would like to see if there are differences between:
- correct/incorrect words
- words that were included/not included
- dyslexic children/control group.

I wanted to be sure that my pipeline was correct, as I am facing some
difficulties.

1) I don't have any electrodes marked as eog or emg. I suppose I can just
rename the two electrodes close to the eyes and the two two at the level of
the cheekbones?
2) Some subjects do not have a reference. Can I procede in this way?

>
> if len(raw.info['chs']) == 130:
>     ref_channels=[]
>     print ('A reference is already present.')
> else:
>     print('Data need to be referenced')
>     raw.set_eeg_reference('average', projection=True)  # set EEG average
> reference
>     ref_channels=['Cz']
>     #raw.apply_proj() #To apply the projection


 3) What parameters should I pass to ICA to remove blinks/movement
automatically? I can do ICA after epoching and then I visually check my
data again, manually selecting other artifacts that have not been detected?

4) I want to implement autoreject in my pipeline. When should I use it?

5) Is it correct to check bad channels visually and using *RANSAC algorithm
from autoreject* over all the epochs? I also tried to check for outliers
using the function *is_outlier* from here https://stackoverflow.com/
questions/22354094/pythonic-way-of-detecting-outliers-in-
one-dimensional-observation-data. It gives me a clue about which electrodes
are different from the others, am I correct? Often times the ransac
algorithm and the is_outlier function gives me a similar list of electrodes.

6)What if I want to use different tmax and tmin for different events when
creating epochs?

I want to be sure about my pipeline as this is the first eeg analysis I am
doing.
- I impor raw data -> I mark bad channels -> I do the reference -> I create
events and epochs -> I run ICA
At this point I actually don't understand how to remove artifacts with ICA.
I just need to create EOG epochs and reject them?

After ICA I visually check the data again and manually remove other
artifacts.

-> Then I do the ERP for the events I am interested in.

At this point I would like to know how I can make the statistics to see if
there are differences between conditions and between groups.


Suggestions?
Kind Regards.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.nmr.mgh.harvard.edu/pipermail/mne_analysis/attachments/20180404/5f29807a/attachment.html