Dear all,
I have run post-stats (Voxel thresholding after pre-threshold masking) with FEAT 5.98 on the same data set under two different operating systems (opensuse vs ubuntu 12.04) and get discrepant results.
After searching for the origin for this discrepancy, I found that the GRF thresholding seems to be calculated differently, thereby providing different results.
Under opensuse the report_log.html shows that the ptoz finds the Z corresponding to zero resels. Because my understanding is that the number of resels for ptoz is calculated as #resels= VOLUME/RESELS, the number of resels should be 11/56.0059=0.196407878. Hence here the value seems to be rounded off (although you get the same result with 0.196407878):
echo 11 > thresh_zstat1.vol
zstat1: DLH=0.0883917 VOLUME=11 RESELS=56.0059
/usr/local/fsl/bin/ptoz 0.05 -g 0
0
(My interpretation of this result is that the ROIs of the pre-threshold mask are so small that effectively no thresholding is carried out according to GRF.)
Under ubuntu 12.04 the exact same analysis on the exact same data (same FEAT version) gives the following result.
echo 11 > thresh_zstat1.vol
zstat1: DLH=0.0883917 VOLUME=11 RESELS=56.0059
/usr/share/fsl/4.1/bin/ptoz 0.05 -g 1
2.000748
Here the number of resels for the GRF thresholding seems to be 1 as if the value of 0.196407878 had been rounded up.
More searches on this topic on the Internet have revealed,unfortunately, no further explanations as to why the correction seems to be carried out differently. Any help regarding an explanation for these discrepant results or a hint which results are more accurate to go with is greatly appreciated.
Thanks in advance and warm regards,
Stephan
|