Generative Artificial Intelligence in Medical Imaging: Foundations, Progress, and Clinical Translation.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Generative AI, including GANs, VAEs, and diffusion models, is transforming medical imaging capabilities.
• Applications span data synthesis, image enhancement, modality translation, and spatiotemporal modeling.
• Generative models address challenges like data scarcity and cross-modality integration.

Purpose of the Study:

• To provide a comprehensive review of recent advances in generative AI for medical imaging.
• To evaluate the expanding roles of generative AI across the clinical imaging workflow.
• To propose an evaluation framework and identify deployment challenges.

Main Methods:

• Systematic review of generative modeling techniques (GANs, VAEs, diffusion models, foundation architectures).
• Examination of AI applications in imaging acquisition, reconstruction, synthesis, diagnostics, planning, and prognosis.
• Proposal of a 3-tiered evaluation framework (pixel, feature, task levels).

Main Results:

• Generative AI enhances various stages of the medical imaging workflow, from acquisition to prognosis.
• The review identifies key obstacles including generalization, data privacy, and regulatory hurdles.
• Convergence with foundation models promises scalable and integrated imaging systems.

Conclusions:

• Generative AI offers significant potential to advance medical imaging, addressing critical challenges.
• A robust evaluation framework is crucial for clinical translation and ensuring reliability.
• Future research should focus on overcoming deployment barriers and fostering interdisciplinary collaboration.