I have around 2000 MRI images with different shapes. I want to map all of them to the same real space (mm space) and unify their shape. So, I picked an MRI image at random and I choose its shape and affine transform as a basis for the subsequent analysis.
For each image in the dataset I have, I dd the following:
new_nifti = image.resample_img(old_nifti, target_affine=basis_affine, target_shape=basis_shape,
interpolation='nearest')
My approach sounds reasonable to me. However, I am afraid of subtleties that I am not aware of. For instance, I am afraid I will end up with something like this. It is hard to keep track for all the 2000 images to ensure their resampled version looks good and reasonable.
- Do you think my approach make sense? Are there subtleties that I did not pay attention to?
- Is there a better approach for choosing the base shape of the MRI image instead of random. The same question also applies on choosing the base affine transform.