Opinions/Rules on which "Scout Function" to use?

Hello everybody,

I'm currently extracting scout time series using an atlas and I have the following option to obtain a single time series for each scout:

image

and I was wondering if there was some information on which function is the "best"/"most appropriate". I personally was leaning towards the PCA function, but saw that some first PCA components explain less then half the variance of my data, so now I'm leaning more towards the Mean function.

I was also wondering if it might be interesting to implement an updated PCA method where you could specify a certain percentage of variance that needs to be explained and that it would keep adding PCA components until that percentage was reached. Just an idea though.

I am curious about your opinions.

and I was wondering if there was some information on which function is the "best"/"most appropriate". I personally was leaning towards the PCA function, but saw that some first PCA components explain less then half the variance of my data, so now I'm leaning more towards the Mean function.

We recommend using a simple average (option "Mean"), as it is easier to intuitively understand what it produces.
The PCA could be a good option as well, but never tested it formally against the mean, therefore we can't give any clear recommendation on why using one or the other. If you are interested in comparing them, please share your conclusions with us.

Note that we fixed the PCA code a few weeks ago, please make sure that your installation of Brainstorm is up to date.

I was also wondering if it might be interesting to implement an updated PCA method where you could specify a certain percentage of variance that needs to be explained and that it would keep adding PCA components until that percentage was reached. Just an idea though.

What we need here is a function that returns only one signal for each ROI. If we keep multiple components for each scout, this would become unmanageable from the interface.

However, if you are interested in testing alternative options, you should be able to find your way through the data structures with the help of the Scripting tutorial, please let me know if there is any information that you can't find.
https://neuroimage.usc.edu/brainstorm/Tutorials/Scripting