A framework for synthesizing contrast-enhanced breast MRI from pre-contrast images using conditional diffusion models. Implements subtraction-based and postcontrast DDPM variants with tumor-aware loss functions and segmentation-guided conditioning for enhanced lesion fidelity