Hi Dr. Jenkinson,
Thank you so much for your reply! I used to use diffunpack in TrackVis to convert dicom into nifti. It generated a nifti file with a different image orientation (I mean it is still LAS, but in fslview, it is upside down compared to how the traditional way an LAS image is showed). So I had to reorient it in order for flirt to work (I tried without doing anything, but flirt gave me a totally wrong result). And my confusion started from there. But now I am using dicom2nii. I don't have this orientation problem anymore. I make sure that all the images in radiological orientation before flirt to avoid future problems.
I just wanted to ask you a few more questions if you don't mind:
1. In general, does flirt change how the data in the result image (the .img file) is stored to the orientation of the reference image? Let's assume the reference is LAS and source is RAS. I know flirt copies the reference sform to the new file's sform. I am more concerned about how the data is stored. The reason is that my lab non-linear registration software only takes the resulting .img and the reference .img for further processing.
2. The voxel index from img2imgcoord is what I see from fslview, right? It means, x=0 is at the left of the image (the right of the patient) in LAS. But is the coordinate system from img2imgcoord just voxel index*pixel dimension or it is the physical coordinates I see in fslview?
3. It is probably a dumb question. I thought that the transformation matrix times the old voxel index would give me the new voxel index. But it wasn't. So what is the mathematical relationship between the old voxel index, the matrix, and the new index? I also thought that it depended on the interpolation method, but img2imgcoord doesn't require that piece of information, so I am not sure how it calculates the coordinates.
Thanks,
Yan
|