questions > ImagePositionPatient, ImageOrientationPatient
Showing 1-2 of 2 posts
Results per page:
Jun 3, 2015  04:06 AM | alexander kraskov
ImagePositionPatient, ImageOrientationPatient

does the dcm2nii take into account ImagePositionPatient, ImageOrientationPatient fields of the DICOM header?

For example, if I acquire one after another (ie physical position of the subject relative to the scanner is identical) two images (eg T1 not diffusion) of the same subject but using rotated Field of View, will dcm2nii produce identical nifti file on the output?

Best wishes
Jun 3, 2015  06:06 AM | Chris Rorden
RE: ImagePositionPatient, ImageOrientationPatient
Both dcm2nii and dcm2niix store a image orientation and translation in the NIfTI header - storing as both a matrix (sform) and quaternion (qform).

You can see this using FSL's fslhd or opening the image with MRIcron and choosing Window/Information file menu item and then selecting the 'reorient' tab.
I would point out that when you rotate an MRI image the acquired image will not be identical - issues like scanner shim, phase/read/slice encoding mean that there will be subtle differences. So even the same sequence will generate subtle differences in intensity and spatial distortion (and in any case random noise and head movement mean that two images with identical orientation and positioning will have slightly different image intensities). Therefore, I would say that the image transformation information can help viewing software know the orientation and provide coregistration routines with a very accurate starting estimate for precise alignment. How those tools use that information varies (for example, with MRIcron if you choose Help/Preferences note you can show the raw data, rotate to the nearest orthogonal plane or reslice the data).

As a very advanced note, I would also point out that dcm2nii sets both the sform and qform to the same value, but any subsequent coregistration might make them different - for example SPM (smartly) only updates the sform: so the qform indicates scanner position and the sform indicates coregistered position. Be aware that most tools (SPM, MRIcron, etc) give precedence to the sform, but for historical reasons ITK-based tools give precedence to the qform. This means if you use a tool like SPM to coregister an image and then use ANTS on the image ANTS will effectively ignore SPM's estimate (last I checked).