How sform is calculate? about dcm2niix HOT 2 CLOSED

This question is out of scope for dcm2niix. dcm2niix uses nifti_clib for these transforms, so that would be the appropriate place for questions. In particular, see nifti_quatern_to_mat44

In brief:

• The NIfTI header includes two independent spatial transforms: the Matrix-based SForm (12 degrees of freedom: shear, zoom, rotation, translation each in three dimensions) and the Quaternion-based QForm (9 DoG: zoom, rotation, translation in 3D). In general, both initially map the image to native scanner space, and a co-registration can update one or both to warp the image to standard space. You can use the sform_code/q form_code to determine the spatial reference (e.g. native vs standard) of each, but most tools give precedence to the SForm.
• Your R-matrix is only 3x3, but you need a 4x4 matrix to include the translations. You must include the qoffset_x/y/z.
• You can easily convert a matrix to a quaternion. However, this transform will be lossy if your transform includes a shear (CT gantry tilt).
• Note you can losslessly convert a quaternion to a 4x4 matrix.
• Note that for 3D acquisitions (but not 2D EPI sequences), dcm2niix will losslessly reorient sagittal and coronal slices to be axial slices, with the spatial transforms rotated appropriately.

from dcm2niix.

Sorry for use the wrong site to post my doubt but thanks for your tips and explanations, they help me a lot.

from dcm2niix.