Hello!
We have a dataset of about 130 subjects. Each have one raw T1 image. These T1-images have been through minor prepping steps (e.g., cutting out neck, centering).
There are about 60+ patients and 60+ controls.
The data has been run through Freesurfer´s recon-all function, and we summarized the volumes of the left and right hippocampus from the aseg output files. We ran wilcoxon unpaired (1-sided and 2-sides) tests, and the difference turned out significant between the two groups. Hippocampus volume smaller in patients. This study was published, and the difference of the median of the groups was about 7489.9 (HC) - 7315.3 (patients) = 175 mm^3.
We then ran the exact same data (the T1 data we used as input to freesurfer) as input to the FSL VBM pipeline. We made the template balanced between the two groups, given that one group had about 5 more subjects.
I was expecting VBM to also find significant differences near hippocampus. Or at least borderline, since VBM might be more conservative since we test so many voxels instead of just one volume value per subject.
But the tfce output from randomize is not significant, not even close (1-p = 0.5/06 as best). We tested multiple levels of smoothing and preformed the tests on both GM, GM-mod and on the jacobians (if the difference was purely volumetric / needed expansion in the patient group). As well as 3 different skull-stripping methods (bet -n, bet -b, SSwarper in AFNI).
Can someone explain to me if this is "expected"?
Or how this can be?
Best of regards,
Robin
########################################################################
To unsubscribe from the FSL list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/WA-JISC.exe?SUBED1=FSL&A=1
This message was issued to members of www.jiscmail.ac.uk/FSL, a mailing list hosted by www.jiscmail.ac.uk, terms & conditions are available at https://www.jiscmail.ac.uk/policyandsecurity/
|