Extract Scouts Max Value from Power Spectra?

Discussions
1

Hello everyone!

I have 2 experimental conditions for which I computed the Power Spectrum averaged in 9 frequency bands so that now I have every subject’s PSD for each condition (e.g.: PSD_S1_C1). I used to run usual paired t-tests between these conditions using Process2, but I have 15002 vertices so, when I corrected for multiple comparisons, I was way too conservative. I thought I may Extract each Scouts maximal value as to run the t-tests on the max value of the scouts and not on single vertices in order to decrease the number of comparisons.

The problem is: when I move each PSD in Process1 and use “Extract Scouts time series” I have as a result a matrix that:

  • create the same matrix whether or not I choose the “concatenate output in one unique matrix” (if I untick this option, the result matrix doesn’t change)

  • this matrix I really don’t understand how can be used for further statistical analysis: I understand that for each ROI I have 9 raws corresponding to the max value that each ROI has in each freq band I chose. In this way, if I have an atlas with 360 scouts, I will end up with 3240 raws in the field “Value”.

  • I can’t project these max values on the cortex, if I right click on the matrix I can only choose Display as time series / Display as image/ Display as table. How can I project these max values per scout on the cortex?

  • Even if I specified in the function “Extract scouts time series” to extract the “max” value for each scout, in the field “scout” of the structure I obtain, it says that the function used was “mean” and not “max” ?

How can I use these matrices of max values for statistical analysis so that I can compute t-tests between conditions? How can I project the average (across subjects) max values of the scouts in each condition and the result of their difference on the cortex.

Is it possible?

Thank you,

Arianna

2

create the same matrix whether or not I choose the “concatenate output in one unique matrix” (if I untick this option, the result matrix doesn’t change)

On a single file, it doesn’t change anything.
On multiple files, it creates or multiple files in output.

this matrix I really don’t understand how can be used for further statistical analysis: I understand that for each ROI I have 9 raws corresponding to the max value that each ROI has in each freq band I chose. In this way, if I have an atlas with 360 scouts, I will end up with 3240 raws in the field “Value”.

Indeed, in output, you get a field “Value” with all the values: Nfrequencies x Nscouts x Nfiles.
The meaning of each value is documented in the field “Description” of the file.

I can’t project these max values on the cortex, if I right click on the matrix I can only choose Display as time series / Display as image/ Display as table. How can I project these max values per scout on the cortex?

It is possible to project scout values on the cortex using the process “Simulate recordings from scout”, but it works only in time domain, it hasn’t been extended to frequency results yet.
http://neuroimage.usc.edu/brainstorm/Tutorials/Statistics?highlight=(simulate)#Convert_statistic_results_to_regular_files
So no, it is not possible to visualize all these results on the cortex.

This approach should be used when you have an strong hypothesis on what are your regions of interest. Draw them or use a few regions from an existing atlas, and this approach will become a lot more manageable.

Even if I specified in the function “Extract scouts time series” to extract the “max” value for each scout, in the field “scout” of the structure I obtain, it says that the function used was “mean” and not “max” ?

The “mean” field you see correspond to what is available as the default in the interface for each scout. This function can be overridden in some processes (like “Extract scouts time series”) but it doesn’t update this value. It could, your question is justified. But it has no impact on any following process anyway.