Dear Theodor,
>
> We are currently working on an older DTI dataset that was acquired with 60 directions and comprises 6 b0's with only one PE-polarity (A>>>P). Therefore we can neither use TOPUP, nor do we have a field map to correct for susceptibility-induced distortions. So far we "directly" run the eddy-current correction followed by DTIFIT. The resulting FA images look in total much noisier than what we are used to and show a lot of hyperintense voxels (intensities above 1) within the brain (see image following this link:
>
> https://onedrive.live.com/redir?resid=96E4D997463AC1C1!2843&authkey=!AJRylBifRFr4TcE&v=3&ithint=photo%2cjpg
what you are seeing is cause by negative values in some voxels in some volumes caused by the spline interpolation. What we often do is to set those values to either zero, or some very small positive value, prior to further processing.
When you say they look noisier than you are used to, are you used to using eddy_correct? The images corrected by eddy_correct are quite noticeably smoothed by the interpolation, and would hence result in images that look (and are) less noisy.
Jesper
|