Hello,
I have a question about the EEG frequency analysis... To run a time frequency analysis is it better to process "recordings" ? Or "sources" ?
Thanks in advance !
Hello,
I have a question about the EEG frequency analysis... To run a time frequency analysis is it better to process "recordings" ? Or "sources" ?
Thanks in advance !
This question could be asked not only for frequency analysis, but for all the possible EEG analysis (ERP, connectivity, time-frequency, decoding...)
The maximum amount of information you can get is already present in your raw/unprocessed EEG recordings. All the processing steps applied on the recordings (frequency filters, artifact rejection, ICA, source modeling...) will reduce the amount information in your signals. ICA and MNE source modeling both produce new signals that are linear recombination of the EEG signals, that can help separate different independent brain processes to study them separately, but that do not add any frequency information to the signals.
The question of whether it is better to work in sensor space or in source space depends on what you are expecting to do with the signals. If you don't need source separation or source localization, then ICA or source estimation might be unnecessary or even detrimental to your analysis.
Whenever one or a few electrodes are supposedly carrying all the information you need in your analysis then you may not need any source analysis (e.g. the amount of alpha power in V1 might be easier to study from one or a small cluster of occipital electrodes rather than from a V1 ROI extracted from a full brain source reconstruction).
See this article: EEG is better left alone
https://www.nature.com/articles/s41598-023-27528-0
Thank you for this very interesting, useful and complete answer!