I am a new user and I have made all the steps to calculate the induced activity of the alpha band with brainstorm. This approach consists in 1) the filtering of the EEG signal in the desired band; 2) rectifying and 3) averaging. This is called Temporal Spectral Evolution (described by Hari).
Then you subtract from TSE modulation, the evoked activity (frequency components of ERPs). As a result of this, you get the induced activity (non-phase).
My question is if this processing of the signal and its result, is adequate to calculate sources with sLORETA? I have tried, and sources look fine but I would like to be sure that we are not violating any assumption.
If you want to get time-frequency decompositions in source space, you need to compute the sources first, and then filter/rectify/normalize/average.
The rectified or normalized signals are not suitable for source estimation with minimum norm estimates, you need real EEG or MEG signals. The only possible step that you could do at the sensor level first is the frequency filtering, but then you need to use signals filtered exactly in the same way to compute the noise covariance.
Thank you very much for your recommendation. I have applied a script with all the steps over the sources of each trial. The result is a high demanding process for the computer. I have reduced even the number of vertices (6000) and overwriting previous files. But still, just one subject (144 trials) is more than 50 Gb.
Use only a few regions of interest, or one frequency band, or average the results online instead of saving all the trials (similarly to what is done by the time-frequency processes of Brainstorm).
In our last contact, you recommend to me that I have to use sources to perform our TSE method, that is, filtering, rectifying (absolute) and averaging (and a baseline correction).
It looks that it works fine when I analyze these averages. However, when I try to perform statistical analysis I receive a message in the command window that says "All values are null, review your input files". Do you know what is wrong with our data?
I have a hint that maybe can help you out. I have made a pipeline protocol to perform the analysis and a warning message says that it was not possible to overwrite links. I have included in all the steps that I described above, always overwrite previous file (just for memory capacity issues). Could it be that the problem?
There is nothing significant in your difference. Set the p-value threshold to 1 see all the t-statistics.
I have made a pipeline protocol to perform the analysis and a warning message says that it was not possible to overwrite links. I have included in all the steps that I described above, always overwrite previous file (just for memory capacity issues). Could it be that the problem?
No. If Brainstorm can't overwrite the input files, it just creates new ones, and keeps going.