Hi all. I'm currently trying to expand upon an analysis in which I only used a small selection of channels as my ROI, and now I'm considering a data driven approach. Is there a consensus on the best method for this, or is it more of a trial and error item?
Initially I thought this was an EEG question instead of fMRI, now I'm not sure... Anyway...
In recent work (attached) we applied singular value decomposition (SVD) to the channel average over all trials (the ERP at each channel) and then applied the resulting channel weights on the raw EEG to obtain single-trial estimates of the ERPs with better signal to noise ratios. Although specially we were looking at the event-locked P200s and N200s, similar methods may also prove useful for looking for response-locked ERPs.
One could also apply SVDs / principal component analysis (PCA) to the channel average as a form of data reduction or component extraction. I recently found a review of those methods by Kayser and Tenke (attached).
Article Consensus on PCA for ERP data, and sensibility of unrestrict...
Article How attention influences perceptual decision making: Single-...
Thanks so much for your answer. Not sure why it came up as fmri.
Very useful, I shall look into it. I've been experimenting with component clustering, but not sure if that is over kill for the components I'm looking at, although on a theoretical level it makes good sense to use.
Are there any tool boxes available for SVD, or is it a case of apply it by hand?
MATLAB, R, and NumPy Python all have svd() functions that call similar linear algebra libraries. An example script/function to calculate single-trial ERP estimates in R is attached.
For data reduction, one could submit the time*channels trial-averaged EEG matrix to a PCA algorithm or just apply the SVD to the trial-averaged EEG and right-multiply reduced U to reduced S (see the attached wikipedia article)
https://github.com/mdnunez/mcntoolbox/blob/master/nunez_etal2016_mathpsych/EEG_R/svderp.R
https://en.wikipedia.org/wiki/Principal_component_analysis#Singular_value_decomposition
- Badges
- Science topic
- Similar topics
- Biological Science
- Physiology
- Biosignals
- Biomedical Signal Processing