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 <img width="826" alt="Screenshot 2024-11-27 at 11 49 30" src="https://github.com/user-attachments/assets/6d9a89e0-5695-4f8d-b1e0-1551efc30a3e">