Resampling: Ghost slices due to rounding errors

[robertschweizer]

I am using dicom2nifti with resampling, to convert DICOM datasets with missing slices to NIfTI.

For certain z-spacings, the resulting NIfTI files have ghost slices at the end of the 3d array, which only contain 0s.

I found that the float32 header.get_zooms() value
when calling resample_nifti_images() leads to rounding errors.
Currently, resample_nifti_images() creates a ghost slice at the end of the nifti volume in this case.

Example code that triggers this problem:

z_spacing = 0.45  # In float64: 0.45000000000000001
nifti = Nifti1Image(
    dataobj=numpy.ones((3, 3, 3)),
    affine=numpy.array(
        [
            [1, 0, 0, 0],
            [0, 1, 0, 0],
            [0, 0, z_spacing, 0],
            [0, 0, 0, 1],
        ]
    ),
)
voxel_size = nifti.header.get_zooms()  # Rounds to float32: 0.44999998807907104492
nifti_out = resample_nifti_images([nifti], voxel_size)
assert nifti_out.shape == (3, 3, 3)  # Without the fix, this is (3, 3, 4)

I'm adding two commits here. Each one would prevent the ghost slices on its own, feel free to merge only one of them.