[FieldTrip] ICA on each trial or collapsing over all trials

Dorothee Hoppe dorothee-brigitte.hoppe at student.uni-tuebingen.de
Thu Apr 2 15:44:05 CEST 2015

Dear Discussion Group members,

we want to use ICA for artifact correction, and now got stuck at the  
point of deciding  whether to compute the ICs on the trial level or on  
the whole dataset for a subject. Using the whole dataset, the  
advantage is probably a higher power of the ICA because of the larger  
data input, so for example a blink component gets probably more  
distinctive. On the other hand, we observed, that artifacts that occur  
only in a small set of trials, like strong muscle artifacts cannot be  
detected that way, or even lead to a distortion of the components for  
that special trial.

Would you recommend an ICA on each trial or collapsing over all trials  
(both options within each subject)? Are there other arguments to favor  
or reject these options?

Best regards,

More information about the fieldtrip mailing list