Abstract
Physics-inspired generative models, in particular diffusion and Poisson flowmodels, enhance Bayesian methods and promise great utilities in medicalimaging. This review examines the transformative role of such generativemethods. First, a variety of physics-inspired generative models, includingDenoising Diffusion Probabilistic Models (DDPM), Score-based Diffusion Models,and Poisson Flow Generative Models (PFGM and PFGM++), are revisited, with anemphasis on their accuracy, robustness as well as acceleration. Then, majorapplications of physics-inspired generative models in medical imaging arepresented, comprising image reconstruction, image generation, and imageanalysis. Finally, future research directions are brainstormed, includingunification of physics-inspired generative models, integration withvision-language models (VLMs),and potential novel applications of generativemodels. Since the development of generative methods has been rapid, this reviewwill hopefully give peers and learners a timely snapshot of this new family ofphysics-driven generative models and help capitalize their enormous potentialfor medical imaging.