Hi Anderson,
I highly appreciate your great feedback.
Actually my aim is to evaluate the relationship between FA and white matter volume in the areas of significant difference between the groups ( two groups ( autism and controls)) and I don't know what is the optimal way to do it! That's why I invented this weird approach to use the statistical maps of TBSS as mask to calculate volume.
Kindly do you suggest me better way to study this relationship?
Many thanks,
jon
Hi Jon,
This analysis isn't appropriate on various grounds I'm afraid.
First thing is to ask what you'll do with the volume data. If a new comparison using the same subjects, there is a serious risk of circularity. See the Vul et al (2009) paper on this matter (a.k.a. "voodoo" paper).
The second aspect is the use of the affine matrix to put the masked significant result from standard to subject space. The volume will be directly proportional to the scaling factor between MNI and the subject space, which can be computed simply as the determinant of the matrix, being the same throughout the brain and (in practical terms) unrelated to the significant regions. Instead, it is an estimate of the total intracranial volume. So, not only it isn't needed to put the region back to standard space using affine, but doing so will actually be giving a measure of intracranial volume, not of the white matter volume.
A third issue is the use of the T1. Although from the command line you don't use the actual information from inside these images, it seems your intention would be to compute a summary quantity from the T1 masked by the significant region. This quantity is unrelated to WM volume.
Another problem is that the TBSS uses a thin skeleton, and even if you warped (i.e., non-linearly) back to the subject space, it would still cover just a small volume of the white matter.
A fifth problem is that WM volume doesn't provide useful information, particularly given that you have much more and better information provided by the diffusion data. Try instead to keep your focus on the diffusion imaging and use as much information as possible from these images. They tell a lot more about WM than the T1 or its derivatives.
All the best,
Anderson
On 14 October 2015 at 16:24, Jon Anderson <[log in to unmask]> wrote:
Hi FSL experts,
Is it correct to use the statistical maps ( tbss_tfce_corrp_tstats1, tbss_tfce_corrp_tstats2) from the output of TBSS analysis as mask to do further calculations.
For example if I want to calculate the white matter volume in the areas of significant difference between the groups in FA ( the output of TBSS) :
When finish TBSS analysis :
1. I threshold tbss_tfce_corrp_tstats2 at 0.95 to get tbss_tfce_corrp_tstats2_thr_0.95
2. I move this thresholded map to T1 ( by registering T1 to MNI and then using reversed matrix)
3. Calculate the white matter volume using the following command fslstats T1 -k tbss_tfce_corrp_tstats2_thr_0.95 -V
Is this correct?
If so do I need to calculate the volume over all the voxels using this command :
fslstats T1 -k tbss_tfce_corrp_tstats2_thr_0.95 -M -V | awk '{ print $1 * $3 }'
Thanks for any advice!
Jon
