Hi - also, MJ has noted that it would be wise to take the abs() of the output if you use higher-order interpolation, just to make sure that all output is positive.
This is easy with fslmaths.
Also, if you wanted faster sinc-like interpolation, you could use the spline option via a call to applywarp (fed from the FLIRT-output transform).
Cheers.
On 14 Mar 2012, at 08:04, Stephen Smith wrote:
Hi - we've discussed this and think we probably agree with you that sinc may be better.
Cheers.
On 14 Mar 2012, at 07:47, Ed Gronenschild wrote:
Hi,
Eddy_correct uses flirt for registration to the reference volume.
By default the final interpolation method is trilinear. I have
used sinc by using the options -sinc -sincwindow hanning.
The result looks more crispy, less blurred, as expected. My question is:
is there any fundamental objection against using sinc interpolation
in the eddy current correction?
Regards,
Ed