Stat maps

Submitted by alex on
 Hi all,
This may seem like a basic question, but I just wanted to check whether entering the single-subject FC or zFCmaps produced by REST  into a one-sample t-test in spm is an appropriate way to generate a group-averaged stat map.
Thanks,
Alex 

YAN Chao-Gan

Thu, 07/02/2009 - 23:58

The correlation coefficient maps (FCmap) were converted into z maps (zFCmap) by Fisher's r-to-z transform to improve the normality. You can find it in lots of statistical books. In my opinion, zFCmaps can enter into SPM's t-tests. You can find lots of such processing in Functional Connectivity studies.
Hope experts of Statistics give some more deep comments.

alex

Fri, 07/03/2009 - 12:45

In reply to by YAN Chao-Gan

 Ok, thanks. I thought as much.

In that case, there's a few other things I would like to check as I am getting somewhat strange results. In the first instance, I am simply trying to generate a map of the default mode network, using a posterior cingulate seed region. I have included motion parameters, wm and csf signal and global signal as covariables. However, when I view the group averaged map in SPM, every voxel in the brain is positively correlated with the seed, and no voxels show negative (anti) correlations. This is what one would expect if the analysis were run with no covariables, so I am wondering whether these were taken into account when I ran REST.

In setting-up the analysis, I loaded the directories containing the detrended, filtered data for my 58 subjects. For the covariables file, I entered the name of a text file, which contained a list in the following format:
Covariables_List:
/path/to/file/subj1_confounds_text_file.txt
/path/to/file/subj2_confounds_text_file.txt
...
/path/to/file/subj58_confounds_text_file.txt

Each subject's confounds_text_file is a data matrix with N rows and 9 columns, where N = the number of time points, and 9 corresponds to the number of nuisance covariables (global signal, wm signal, csf signal, + 6 motion parameters).

Does this sound correct? The analysis seemed to progress to completion without any problems, its just that the final maps do not look like what one would expect.

Also, is there a way to read in multiple text files for the seed timeseries? I can only load one text file in the Time courses option of the ROI definition gui, which suggests that only one subject can be analysed at a time if the seed timeseries are loaded in from text files.

Thanks again for your help,
Alex

YAN Chao-Gan

Fri, 07/03/2009 - 23:16

In reply to by alex

What's the order of your subjects' directory?



If you had the order of subjects' directory as shown in the above image, you need to arrange the covarible list file as following:
Covariables_List:
X:\Structural_Functional\Reprocess\Sub_428_Cov.txt
X:\Structural_Functional\Reprocess\Sub_427_Cov.txt
X:\Structural_Functional\Reprocess\Sub_426_Cov.txt
X:\Structural_Functional\Reprocess\Sub_425_Cov.txt
X:\Structural_Functional\Reprocess\Sub_424_Cov.txt
X:\Structural_Functional\Reprocess\Sub_423_Cov.txt
X:\Structural_Functional\Reprocess\Sub_422_Cov.txt
X:\Structural_Functional\Reprocess\Sub_421_Cov.txt
........

I am sorry I did not tell you before, and I think there maybe something wrong with your covarible list file. Thanks for your report, we will and a message box to tell you how to arrage the covarible list file in details after you click select covariable file button.


Currently, you can only load one text file in the Time courses option of the ROI definition gui, which suggests that only one subject can be analysed at a time if the seed time series are loaded in from text files. Or you can try command line of fc.m.
We will and this function to the GUI in the next release, we will and a couple of convenient functions to the next release, please be a little patient. Thank you very much!

Forums