 |
 |
Hi, If the registration is working well then different options should give very similar results. If you are seeing substantially different results then that probably indicates that one or other (or both) of the registrations is performing badly. The options that we use in eddy_correct are there to make things a little more robust, and often it is fine to go with different options. However, you can never tell with registrations when things might not work well, so if you are seeing these differences then I would look carefully at the eddy_correct outputs and try to see which is doing better and whether there is still room for improvement. I hope this helps. All the best, Mark On 3 Apr 2012, at 18:28, Yiou Li wrote: > Hi Mark, > > Thanks for the clarification. But my question is why those two options are used in eddy_correction. --- Are those options used to improve the estimation of the diffusion tensor parameters? --- I find that some people use the default flirt command without using any options to preprocess the DTI scans so I tried that way and found that the estimated FA maps are different from the one using eddy_correct. So I wonder which method should I use. > > Best, > Leo > > > On Tue, Apr 3, 2012 at 6:50 AM, Mark Jenkinson <[log in to unmask]> wrote: > Dear Leo, > > The -nosearch is to stop an *extensive* initial search of rotation angles. > It does not the optimisation method from finding non-zero rotations, and > you will find that it does in fact correct for rotations. > > The -paddingsize is to make sure that end slices are not lost due to small, > half-voxel shifts of the FOV. This just copies the intensities from the > nearest end slice into any areas which have fallen outside the original FOV > by up to 1 voxel. > > All the best, > Mark > > > > > On 2 Apr 2012, at 23:27, Yiou Li wrote: > > > Dear FSL experts, > > > > Could you please advise the reason of using the -nosearch and -paddingsize 1 in the flirt command in eddy_correct? > > > > It seems to me that -nosearch constrains (or turn off?) the rotation in registration transformation -- is this because rotation transfomation could introduce artifacts in estimation of diffusion parameters? > > > > Thanks in advance! > > Leo > > > > >