Question about ft_freqanalysis and ft_sourceanalysis

Discussions
1

Hi :slight_smile:

Although these tutorials of are under construction, are they ready for use? When I computed ft_freqanalysis (on the FFT on sensors) and tried to open the file that was created, it was written “vector must be the same length” and the file was empty. Concerning ft_sourceanalysis, the computation (on the FFT on source) did not work, it was written “pow not present in data”.

Is there anything I should know that may be the cause of the problem? What I do is that in process 2, I put all the FFT files from my sujects in condition one in the “file A box” and I put all the FFT files from my subject in condition two in the “file B box”.

Thank you,
Daphné

2

Hi Daphne,

This work in progress, the implementation is changing almost everyday.
What you’ve been trying to do should be working now (stat on the FFT both at the sensor and source level).

Can you update Brainstorm and try again?
If you’re still getting errors please post the full error messages here, together with a screen capture showing what your database looks like.

Thanks
Francois

3

Thank you Francois,

However, it did not seem to change after I updated brainstorm. Here is a screen capture showing the error from ft_sourceanalysis with sources

BS_problem_sources.PNG1069Ă—1037 141 KB

and this one from ft_freqanalysis with the sensors
BS_problem_capteurs.PNG957Ă—1026 103 KB

.
Or maybe the update did not work..

Thanks!
Daphné

4

Hi Daphne,

I identified many issues with what you were trying to do. If think I fixed what I was due to the program.

There are also things to question about the test that you are trying to run:

  1. ft_sourcestatistics is designed to perform tests on full cortex maps, while the files you have in input are scouts. I added an error message that says “use ft_freqstatistics instead”.
  2. If you are planning to use the cluster-based correction, you have to be aware that ft_freqstatistics will not consider two scouts as “neighbors”, so it will not be able to aggregate clusters in space. It will use the only dimension available, the frequency dimension, which is very sparsely sampled because you computed averages of frequency bands from the original PSD results.
  3. It’s fine for running non-paramtric tests, but if you want to use the cluster-based correction, you might prefer to use the full spatial resolution (no scouts) and the full frequency resolution (no frequency bands in the PSD computation). I’m not sure about this last statement, you should double-check this with FieldTrip experts to know how they would deal with this.

Francois

5

PS: start by updating brainstorm before you try anything new, I fixed many things and added the cluster display for the PSD spectrum display.

6

Hi Francois,

Perfect. I updated brainstorm this morning.

Thanks for the specifications. So I tried to run ft_sourcestatistic on the full cortex map. It did not work the first time because the file did not have the same number of time samples

Bs_problem_source_sample1.PNG1065Ă—1036 114 KB

. Howerver, I after reduced the time selection so that it include all the files, it still give the same message but I don’t know why
bs_problem_source_sample2.PNG1128Ă—1034 116 KB

. Am I doing something wrong?

Thank you very much,
Daphné