 |
 |
Hello again, I'm slightly puzzled right now... I registered my volume and the registration matches really well. Previously I had noticed that the volume had flipped around z-axis, so I created a zlip-matrix and gave it as -init for flirt. Ok, as I said, the registration is now really good. But, transforming a certain coordinate doesn't succeed. Voxel sizes are different in the volumes. My transform-matrix is: Mat= [0.997118 -0.075060 -0.004070 12.922070 0.070709 0.998419 -0.030878 12.680435 -0.010383 -0.031842 -0.998170 186.943241 0.000000 0.000000 0.000000 1.000000] I tested the transform for the outermost voxel in the input volume. I have tested both "voxel coordinates" and "mm-coordinates". They would be: coord_vox=[255 255 179 1]' and coord_mm=[220 220 180 1]' I tried both "direct" transform, as suggested, like new=xfm*old, and also one way which was introduced previously in Mark Jenkinsons reply to Souheil in Nov 14. Because I have different FOV:s I thought that might be the answer. That way would be: new=inv(Ddest)*XFM*Dsource*old as I understood. Ddest and Dsource have as diagonal elements the x, y and z voxel size in mm, and 1 as fourth element. Like: D= [ 4 0 0 0 0 4 0 0 0 0 6 0 0 0 0 1 ] My voxel sizes are: Dsource=(0.859 0.859 1) and Ddest=(1.02 1.02 1.5). What am I doing wrong, because all of those transformations gave negative value for z-coordinate? Is the corner point somehow "bad verification"? Any help is greatly appreciated.... With best regards, Pasi T.