I recently upgraded a computer in the lab from an earlier version of FSL to FSL v5.0.10 and wanted to make sure that it produced the same output as a new computer that only had FSL v5.0.10 installed. I ran my DTI script (which includes topup, eddy, and dtifit) on the same participant across computers, using the same raw data file and bval/bvec file inputs. When I looked at the output of dtifit, the FA maps were different.
To confirm that the difference had to do with the version installations, I then reran the same participant on each of the computers. Surprisingly, the output of dtifit was different. So I then compared each output file from each processing step of the first and second analyses of the same participant on the same computer. It turns out that the .eddy_rotated_bvecs output of eddy, which I used as input for dtifit, was the only file that differed. Importantly, the data file output of eddy yielded similar intensity values across the first and second analyses of the same participant. This suggests that the bvec rotation varies each time it is run on the same subject, even on the same computer.
Am I missing something or is this quite concerning?
|