Dear users,
I am trying to compute the Hilbert transformation on a sources file computed with MNE and sLORETA (Unconstr.) on a continuous EEG recording (121 s), using several scouts (Desikan), but the buffering bar is always at 0%.
I tried to follow the tutorial instructions (chapter 24), that is choosing the raw file, but I faced the same problem. Could anyone help me?
Thanks to everyone,
Greetings
You will most likely not be able to process the time-frequency decomposition of all the source signals (45000) over all the file (121s). This requires the reconstruction of gigantic files in memory.
For example: at 500Hz, the full source file would be 45000 x 121 x 500 x 8 / 1024^3 = 20 Gb
The Hilbert transform with 6 frequency bands would need in memory at least: 20 * 4 * 6 = 480 Gb.
The only process optimized for processing continuous files (and maybe the only one that really makes sense in terms of frequency analysis) is the PSD:
https://neuroimage.usc.edu/brainstorm/Tutorials/RestingOmega
You will need to reduce the dimension of your data:
- work with shorter epochs
- use the option "scout function: before" instead of "after":
https://neuroimage.usc.edu/brainstorm/Tutorials/RestingOmega - work with a PSD by large time windows instead of a Hilbert transform (the full time resolution is most of the time not very useful in resting state analysis)
- ...
Thank you very much for your advices, I suspected indeed that the registration file was too long!
Best regards,
Giulia.