MedM2G: Unifying Medical Multi-Modal Generation via Cross-Guided Diffusion with Visual Invariant | Read Paper on Bytez