[Elastix] transformix (interpolation?) issues

Pelt, R.F.P. van R.F.P.v.Pelt at tue.nl
Mon Apr 8 15:01:07 CEST 2013

Hi Marius,

Thanks a lot. Works like a charm!
Figured is was in the interpolation, but did not think of altering the transformation file.

Thanks again.


Van: M.Staring at lumc.nl [M.Staring at lumc.nl]
Verzonden: maandag 8 april 2013 14:56
To: Pelt, R.F.P. van; elastix at bigr.nl
Onderwerp: RE: transformix (interpolation?) issues

Hi Roy,

For binary images you need to manually set the interpolation order to zero.
Copy the resulting TransformParameters.0.txt file and replace the relevant line with:

        (FinalBSplineInterpolationOrder 0)

Regards, Marius

-----Original Message-----
From: elastix-bounces at bigr.nl [mailto:elastix-bounces at bigr.nl] On Behalf Of Pelt, R.F.P. van
Sent: maandag 8 april 2013 14:43
To: elastix at bigr.nl
Subject: [Elastix] transformix (interpolation?) issues

Dear all,

While registration results on MRI imaging data look quite good, I decided to convince myself of the results with a small example. I've created a binary tube as a test case for a simple rigid registration.
The moving image is just rotated by 10 degrees, and the registration in itself performs excellent.
The deformation field is nice and smooth, and deforms in the expected direction.

I'm aware that binary data is not representative for the imaging data.
Also, the mutual information measure is not ideal for the binary case.
However, I'd like to keep my parameter file identical to my real data case, and keep the synthetic extremely simple.

However, transformix seems to give unacceptable (interpolation?) artifacts?
I've attached an overview image and the parameter file.
Any idea what could be going on here?
(Data sets are 128^3 volumes with isotropic scaling, and the tube diameter is about 20 pixels)




More information about the Elastix mailing list