[Eeglablist] Proper Sequence for Manual Rejection, ASR, and ICA in EEG Preprocessing and Related Questions

[Cedric Cannard]

Hi Su,

Congrats on the preprint at Scientific Reports and happy to hear you like BrainBeats and that it is useful. Let me know if you have any questions that are not answered by the documentation or the JOVE paper. 

You can definitely clean your data before running BrainBeats if you have large portions that are bad. For HEP, just make sure you keep both EEG and ECG signals together when you remove bad segments so that they remain time-locked. If they are not already together, I can give you code to put them together. Or you save the latencies of the bad segments you remove one type of signal (e.g., ECG) and remove the corresponding segments from the EEG (EEG = pop_select(EEG, 'nopoint', latencies_of_bad_segments). If you do the cleaning, simply switch off the preprossings in the GUI. But BrainBeats should deal with those automatically, can you describe what's not working for you? R-peaks located in bad segments will not be detected and therefore not included in the HEP data, so you shouldn't have to worry about it. If they are interpolated, turn off interpolation of RR artifacts, but if bad epochs make it through, they should be detected and removed by BrainBeats.   

See Makoto's answer about ICA on epoched data. 

I would exclude EOG channels if ICLabel is removing ocular components well. And see other conversation about ECG. To give us some answer, you could try both: HEP results when removing heart components vs. results when keeping heart components. 
If you want to remove heart components, I also have seen better performance keeping the ECG channel. 

Let us know how it goes. 

Cedric



On Wednesday, August 7th, 2024 at 10:15 AM, Makoto Miyakoshi via eeglablist <eeglablist at sccn.ucsd.edu> wrote:

> Hi Bong,
> 
> > I am concerned about whether ICA can be applied if edges are segmented
> 
> due to rejection. Isn't this the reason for performing epoching after ICA?
> 
> When ICA starts, it shuffles the time points of the input data. For every
> iteration, it shuffles the data differently. So it's not a problem for ICA
> that the input data are completely chopped up and even random-ordered in
> time (here, you can see the meaning of 'instantaneous independence'; ICA
> has no concept of time).
> 
> > including EKG might even remove HEP
> 
> 
> Why/How does it remove HEP?
> 
> Makoto
> 
> 
> On Tue, Aug 6, 2024 at 9:09 AM 봉수현 via eeglablist eeglablist at sccn.ucsd.edu
> 
> wrote:
> 
> > I hope this message finds you well. I am currently in the process of
> > setting up a preprocessing pipeline for my data and have encountered a few
> > questions that I hope you can help me with.
> > 
> > My primary question is regarding the use of Clean_rawdata and where manual
> > rejection should be placed in the sequence—should it be performed before or
> > after applying Clean_rawdata? Additionally, I am curious whether manual
> > rejection should be done before or after ICA.
> > 
> > To provide more context, here is a detailed description of my research
> > problem.
> > I am attempting to epoch my data based on the latency of the 'R-peak'
> > detected in the EKG channel (yes, I am analyzing HEPs). I am considering
> > using various algorithms for this purpose, but the Brainbeats package by
> > Cannard is currently my top choice due to its remarkable effectiveness in
> > my last project.
> > However, I sometimes encounter noise in the EKG channel that obscures the
> > R-peak, which I plan to reject manually. Since the Brainbeats package
> > performs R-peak detection, ICA, and epoching with 'one click,' I am
> > considering removing unnecessary segments before this process. This would
> > include manually rejecting segments where the R-peak is indistinguishable
> > and remove/interpolate frontal lead/leads containing abnormal high
> > amplitude noise occationally (I don't know what's problem with these lead).
> > 
> > Given my intention to include such manual rejection, I am unsure about the
> > optimal sequence for applying clean_rawdata (ASR), manual rejection, and
> > ICA. Specifically, I am concerned about whether ICA can be applied if edges
> > are segmented due to rejection. Isn't this the reason for performing
> > epoching after ICA?
> > Moreover, I would like to know if 'detrending' is necessary for
> > experiments similar to mine. FieldTrip tutorials include detrending, but I
> > am unsure whether it should be done before or after epoching. 
> > Do I need it for my project?
> > 
> > Lastly, regarding the ongoing debate about including EOG and EKG channels
> > in ICA: based on my reading, excluding EOG and EKG channels during ICA does
> > not significantly impact the results and including EKG might even remove
> > HEP (assuming they exist). If there are differing conclusions, I would
> > appreciate hearing them.
> > 
> > Thank you for your time and assistance.
> > Best regards,
> > Bong
> > 
> > Su Hyun Bong, Phd candidate
> > KAIST, GSMSE, Korea, Daejeon
> > Tel 042-350-4285 Mobile 010-9891-8020 Email 
> > npdrbong at kaist.ac.kr mailto:npdrbong at kaist.ac.kr
> > 
> > _______________________________________________
> > Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> > To unsubscribe, send an empty email to
> > eeglablist-unsubscribe at sccn.ucsd.edu
> > For digest mode, send an email with the subject "set digest mime" to
> > eeglablist-request at sccn.ucsd.edu
> 
> _______________________________________________
> Eeglablist page: http://sccn.ucsd.edu/eeglab/eeglabmail.html
> To unsubscribe, send an empty email to eeglablist-unsubscribe at sccn.ucsd.edu
> For digest mode, send an email with the subject "set digest mime" to eeglablist-request at sccn.ucsd.edu