DM-FNet: Unified multimodal medical image fusion via diffusion process-trained encoder-decoder. This paper has been accepted by IEEE Transactions on Multimedia (TMM) in March 2025.