Most upscaling and super-resolution techniques I’ve seen use various implementations of interpolation; typically nearest-neighbor approaches. Although I don’t work in the medical field and haven’t checked in on the research at least since ViTs overtook CNNs for other areas of computer vision.