It's helpful to accelerate especially on large dataset.

For .nii data, can use kvikio + cupy + nibabel
For .dcm data, can use kvikio + cupy + pydicom