Hello. Thanks for such an amazing toolbox. I would like to suggest a new feature for Brainstorm, which is the ability to run individual alpha frequency (IAF) quantification. This analysis consists in determining the maximum power value in a range of frequencies (most commonly alpha) per condition, per subject, which has been demonstrated to be relevant for understanding individual differences in cognition. One recent implementation is described here (https://doi.org/10.1111/psyp.13064). The IAF analysis is becoming more popular now in the fields of multisensory perception and binding. It would be great if you could implement it on Brainstorm.
Thanks so much!
1 Like
@Sylvain @Luc @Marc.Lalancette
Would you be interested in exploring this?
First impression: The method is just smoothing the spectrum and finding the alpha peak frequency. It's not clear from the abstract if the proposed filter for smoothing (Savitzky-Golay filter) is optimal for that task in any way. I'd suggest this can most likely be outperformed by SpecParam/SPRiNT/FOOOF, already in Brainstorm, and probably can be as automated with proper parameters (say one peak max limited to the alpha range). Plus the spectrum model can be statistically tested for the presence of an alpha peak. So I'd say our efforts are probably better spent continuing to improve our current approach.
Probably still worth looking at the paper and references in a little more detail to see if there are technical subtleties of interest, like the filter choice and the peak detection (do they use a specific peak shape?).
Cheers,
Marc
References to methods currently available in Brainstorm:
Thanks so much for the replies. Indeed, the FOOOF method seems superior!
Greetings!, Here's few small questions related to IAF in BS...Any thoughts or suggestions appreciated!
We computed peak frequency in alpha outside of brainstorm and then fed multiple start-stop Hz steps (e.g., for lower and upper alpha sub-bands, etc). We did so by script, making a new set of spectral outputs in BS for each Data Folder in the Database and specifying the subject-specific start and end of each set of Hz Bins. Each of the new sub-bands were given a common name (e.g., Alpha1, Alpha2), but the specific start/stop for that sub-band were unique to each participant (e.g., 8.35 to 10.35 for one subject, and 9.13 to 11.13 for another subject).
Would the easiest way to eliminate the extra (specific per subject) text in the resultant sub-band labels by editing the Labels in the BS spectral output file ? Or might there be a way to turn those off in the BS function itself ?
Is there any useful info from Forum members about successfully taking spec_param outputs from within BS and, with BS GUI, successfully generated IAF-related metrics ? Second, is spec_param expected to work just as well on source time-series as sensor time-series within BS ?
I can probably help with this, but I don't have enough information to start with.
Please post here:
- A screen capture of the figure you get, and a precise description of what you would like to edit.
- The example file used to create the figure (zip it, upload it somewhere, and post the download link here)
Is there any useful info from Forum members about successfully taking spec_param outputs from within BS and, with BS GUI, successfully generated IAF-related metrics ? Second, is spec_param expected to work just as well on source time-series as sensor time-series within BS ?
Will send info as requested thank you for considering
Yes spec_param is meant to be used on sources; the code should work fine.
Since "work just as well" is a bit vague, let me add that one should always keep in mind that the limitations of using an inverse model (resolution, leakage, etc) remain on any subsequent analysis in source space. But nothing special about spec_param in that regards.
Thanks for all the information. I was wondering if there is a way to extract the individual alpha frequency of a PSD file (having used the pwelch method) with the precision that most papers report. They show with 0.1 Hz precision what is the frequency band that contains the highest power value in the frequency band range of interest. For example: https://doi.org/10.1016/j.cub.2014.11.034