Hello,
I would appreciate any criticism/suggestions on the following idea.
In the clinical DTI, I almost always see a subject's motion in the consecutive
images acquired with different orientations. This may come from subject's
rigid motion or eddy current and can be effectively compensated for by
running motion-eddy correction prior to dti_fit and bedpost. I was curious
however, how the motion affects FA, MD and tensor indices, if it is not
corrected for and what is the critical level of motion inducing statistically
significant changes in the above parameters.
Here is what I briefly did:
1. I took a single DTI subject data and corrected for the motion. here I
assume that residual root-mean-square (RMS) is 0
2. in the result of 1. I induced artificial motion by applying a random affine
transform to each volume in the DTI series. Each random transform had RMS
from small interval. I generated several series with the same and increasing
amount of motion
3. then I ran dti_fit on all the data (1. and 2.)
4. merge results of dti_fit from 2 and 1 to the single 4D and ran randomise to
get paired t-test
Here I have a small problem as I am not completely sure, which result file of
the randomise I should look at. This is because I get values above 1 in the
tstat file and very small values (suggesting no significant difference) in
the vox and max files even with high level of motion.
Thanks in advance,
Martin
|