Question about ICA template matching with different number of channels

[yaqing]

Dear List,

I’m looking into removing ICA components in real time, and found that one way to do so is to store a numpy array of template IC derived from existing data, and use the corrmap() function to find components in the current ICs that best correlate with this template.

However, I noticed that the IC array has the length of number of good channels. Given that different acquisition runs often have different bad channels identified, I wonder what would be the best practice to overcome the channel mismatch between the stored template and running data. Should I:

• Interpolate both the data used to generate IC templates and the running data so that both include all channels?
• Interpolate both the template IC and ICs from the running data?
• Interpolate template IC or the data, then remove entries that correspond to bad channels in the running data?
• Something else?

There was a similar question from the archive but seems still open:

512×512 77.5 KB

Hello,    I have a very large dataset where unfortunately we did not have an EOG or ECG channel. Thus, I am interested in using template matching using corrmap. I have a small dataset with lots of robust artifacts I'd like to use to derive the templates. So, I performed an ICA on those, and identified some candidate IC's for templates. Since I'd really rather not load ALL the IC's from my small dataset just to get a handful of template IC's, I'm using the option of giving an array as the te…

Thanks in advance,
Yaqing