Hi Eszter,
Overall the procedure is good, but you can improve it bit. Here are a few comments:
I have loaded and reviewed the continuous data file into Brainstorm, and after applying the visualization filter, I have performed the SSP to correct for EKG and eye artifact.
The visualization filters are for visualization only, they have no impact on any processing step, including the SSP correction.
Next, I have imported this file into Brainstorm as epochs of 5 seconds in length (including DC offset for the entire continuous file).
Depending on your sampling rate, you can probably work with segments that are longer than 5s, it could make it easier to manage to have less files in the database.
Following this, under the processes tab, I have entered the 60 x 5 second files nd have applied the sinusoidal removal (notch).
I would recommend you run the sinusoid removal BEFORE epoching instead. With the CTF .ds and FIF file formats, you can run directly this process on the continuous files (drag and drop the Link to raw file in the Process1 box). Other file formats are not support yet but will be before the end of the year. If you can’t process directly your original files, try importing segments of recordings as long as possible in the Brainstorm database (this will directly depend on the amount of memory you have on your computer - with 32Gb you can import and process the entire file). For visualization, you can then re-epoch the long imported block in smaller epochs.
This process works much better on long time series, and does not perform any correction on the edges of the signals. Therefore, if you run this on the imported segments, you multiply the number of time samples that are not properly processed. To avoid those edge effects, you should always run this process on the longest possible time segments.
Finally, I have reviewed the epoched files individually to remove any “bad” epochs prior to the FFT.
Am I correct in thinking that the next step is to simply enter the epoched files in the Process1 space and under the Frequency option, select the FFT analysis?
If you do this, it will calculate the FFT for each block of 5s, and then average the power of the FFT across those blocks.
The frequency resolution of the final power spectrum depends on the length of the individual blocks (number of frequency bins ~ number of time samples).
Another, simpler way obtain the same thing in a more flexible way: run the PSD process on the continuous: drag and drop the Link to raw file in the Process1 box, then run the process “Power Spectrum Density (Welch)” process with an estimator time window of 5s. An overlap between the different windows allows to increase the number of segments and the stability of the estimator.
The Welch method is a more stable estimator of the power spectrum than the simple FFT. For more information, you can refer for instance to the help of the pwelch method in Matlab: http://www.mathworks.com/help/signal/ref/pwelch.html
To skip the bad segments in the calculation of the PSD, before running the process, review the continuous file (5min) and mark the segments as bad.
Select with your mouse the time segments to ignore and right-click > Reject time segment (CTRL+B).
Later, when you import the files in the database, all the imported trials that contain a piece of BAD event and going to be marker as bad.
Please let me know if you have further questions.
Cheers,
Francois