Recent papers (e.g., Chang et al., 2016) have suggested that eye blinks can be effectively and more efficiently detected using a single channel (e.g., Fp2) as opposed to bipolar channels (VEOG). What do Brainstorm developers think about this and are there plans to add this capability? If one were to only use a single channel for detection in the current version of Brainstorm, what impact does that have on subsequent SSP analyses?
Thanks!
Ann
Hi Ann,
You can do this in Brainstorm: when running the "Detect eye blinks" process, you can select the channel you want. We actually recommend doing this when you do not have an EOG available in your recordings, provided the channel you choose is as close as possible to the eyes.
I'll let others more experienced comment on the impact it has on subsequent analyses.
Cheers,
Martin
Hi Martin,
Thanks for your response. I have used single channels vs. EOG to detect eye blinks in the Brainstomr process and I get different results depending on which channel I select. Importantly, I also get quite different SSP results. In several papers they use different algorithms for detection when using single channels. I was just curious if others have an opinion on this and whether channel selection might call for different algorithms that would improve blink detection/rejection or if it doesn't matter much.
Thanks,
Ann
Hi Ann,
I am not aware of anybody having formally tested in Brainstorm the detection from a single channel (common reference, linked mastoid or average ref) vs a bipolar VEOG channel.
No matter which detection method you use, you should look at what is detected: review manually the events detected with one and the other procedure and try to understand what is detected in one case and not in the other.
If you obtain very different SSP topographies, it is probably because the two signals do not show the same types of events. A single frontral channel (eg. Fp2) recorded with a Cz reference would probably show both saccades and blinks, while a vertical VEOG channel would show mostly blinks and a horizontal HEOG mostly saccades and other eye movements.
In the context of SSP, you would like to get categories of events that are as specific as possible. Only events corresponding to exactly the same type of artifact should be grouped in the same category. If you observe different types of ocular artifacts, you should try to find a way to obtain one category of events for each type of artifact, and then compute separate SSP projectors for each of them.
Francois