Could you provide the whole dataset mentioned in the paper "We collected 3,200 cases to train our diffusion model" to download ? Thanks !