[FieldTrip] [Fieldtrip] Analysis Options for SSVEPs
Schoffelen, J.M. (Jan Mathijs)
janmathijs.schoffelen at donders.ru.nl
Thu Feb 23 12:12:43 CET 2023
This question is not at all dumb, it’s good one!
A classical SSVEP analysis indeed estimates the power of the evoked response (i.e. first average the trials, then compute the power spectrum). This works well under the valid assumption that the timing of the individual stimuli is 1) very precise, and 2) not jittered across trials.
Now, if you want to estimate time-varying changes in the evoked power - e.g. to use it as a ‘marker’ for the attentional focus - you need to keep in mind that the specificity of your spectral estimate (+your sensitivity to detect changes over time) (i.e. the extent to which the estimate at - say - 15 Hz reflects only the energy in the signal at 15 Hz or that it also contains energy of close by frequencies -> a phenomenon that relates to the notion of spectral leakage, and the time-frequency trade off) depends on your experimental stimuli and your analysis parameters. In your example, using frequencies of 13 and 15 Hz, they may not be sufficiently far apart in the spectral domain in order to be sufficiently reliably resolved given your most likely choice sliding time window that you are going to use for your analysis.
Concretely, the kernel that you will use for your time-frequency analysis should have a bandwidth that is small enough to be able to separate the stimulus frequencies. This requires a sliding time window width of at least one second (because a time window of 1 second yields a spectral resolution of 1 Hz), which is probably too long for your experimental design (because it would require trials that are correspondingly much longer than that time window obviously). Therefore, I suspect that you may need to think a bit (and discuss with your co-workers) the optimal stimulation frequencies for both hemispheres.
Another thing to keep in mind w.r.t. the optimal frequencies, is that ideally the lower order harmonic frequencies of the stimuli don’t overlap, because it may be interesting later on to look at the power of the harmonics as well, and in this case those harmonics will still distinguish between stimuli.
Good luck with your experiment,
> On 23 Feb 2023, at 10:35, Nathan Han via fieldtrip <fieldtrip at science.ru.nl> wrote:
> Hi all,
> I'm still quite a novice with Fieldtrip and signal processing so please forgive me if this question is dumb :)
> I'm running an experiment where there is a visual stimulus on the left and right side of the screens, the purpose of which is to elicit SSVEPs. One circle flickers at 13Hz (e.g., the left side) and the other flickers at 15Hz (e.g., the right side). I would like to analyse the change in visual spatial attention over the course of the trial and one way I was thinking was that if attention switches from the left to the right side of the screen, that 13Hz power would reduce while 15Hz power would increase over the course of the trial.
> I'm not sure if that even makes sense or if it is possible. If it does, I would like to ask how should I approach this analysis?
> Kind regards,
> fieldtrip mailing list
More information about the fieldtrip