Application of the CSD to long epochs

Support & Discussions
1

Hello,

I am working with long EEG epochs (from -0.6 to 9 s) and applying CSD (surface Laplacian) transformation using mne.preprocessing.compute_current_source_density(). As I understand it, CSD is calculated as a second-order spatial derivative applied independently to each time point, so in principle, the length of the epoch should not matter.

However, I wonder if there are any known limitations, recommendations, or considerations when applying CSD to such long epochs. In particular:

  • Could long-duration or spatially diffuse generators introduce artifacts patterns into the CSD output?
  • Would you recommend applying the CSD to the entire epoch, or would it be better to segment the data (e.g., into stimulus-locked and response-locked epochs) before calculating the CSD?

Thank you in advance!

Johan

2

Update: I applied both strategies, and as expected, the results were similar. I hope this may be helpful to others!

Best,
Johan