Integrating radiologists' eye gaze patterns and radiomic feature maps as controls to text-to-image diffusion models can generate high-fidelity and clinically accurate medical images.
