Hi Marie,
I don't think so, since your volume does have a small but real rotation away from coronal. Any idea why that is? I guess we could avoid resampling if the direction cosines are close enough to what we want.
Bruce On Fri, 16 Nov 2012, Marie Schaer wrote:
Hi everyone,
I have a question related to mri_convert (I use FreeSurfer version 5.1). I noticed that when I convert some images without specifying anything about the orientation (mri_convert MY_DICOMS 001.mgz), in a high proportion of the case, mri_convert will identify an orientation matrix which is not simply made of 0 and 1, like e.g.
i_ras = (-0.00872651, -0.999962, 0) j_ras = (-0, -0, -1) k_ras = (-0.999962, 0.00872654, 0)
The non integer values result in the application of a slight rotation when converting the images, associated with interpolation and a slight loss in image quality.
The only way that I found to avoid that the image quality loss was to run a first time mri_convert to look at these parameters, and then transform them to the nearest integer to feed them to mri_convert. In the example above I would run: mri_convert MY_DICOMS 001.mgz -iid 0 -1 0 -ijd 0 0 -1 -ikd -1 0 0
Given that I don't change the orientation (i.e no flip), but just want to avoid the rotation, I was wondering if there was any better option in mri_convert than what I am currently doing.
Many thanks in advance for your answer,
Marie _______________________________________________ Freesurfer mailing list Freesurfer@nmr.mgh.harvard.edu https://mail.nmr.mgh.harvard.edu/mailman/listinfo/freesurfer