Dear Theodor,
>
> Thanks a lot for your quick reply! I followed your advice and set the negative values after eddy current correction to 0.0001. Interestingly, this improves the hyperintense, noisy voxels at the edges of the brain. However the hyperintense voxels within the brain so far seem to be unaffacted by this. The FA-maps therefore still look quite noisy.
My advice then would be to extract the “time series” for one of these hyperintense voxels and see if you can figure out what is going on.
> Concerning the eddy current correction, we have always used the eddy_openmp command (we do not have a CUDA-gpu), which, as far as I understood, corresponds to the "newer" eddy command.
Yes.
> On the other hand, we have also used topup in our previous datasets, so I guess the interpolations done within topup also smooth the images usually? As we unfortunately cannot use topup this time (all b0-volumes have been acquired in the same direction), do you think it might be advisable to use eddy_correct for this dataset instead of the new eddy command in order to decrease the noise?
If you use topup+eddy as recommended, i.e. feed the topup field and parameters along with the original data into eddy, there should only be one interpolation regardless of if you use topup or not. So I don’t think that is the reason. I do agree that your FA maps look noisier than I would expect from 60 directions. Do these images have very high resolution (i.e. higher than those you have used topup on) or can there be some other factor that makes them noisier?
I would nor recommend eddy_correct. The tests we have performed indicate that it doesn’t do a good job of correcting for eddy currents/movements.
Jesper
|