Exploring the Applications of Artificial General Intelligence in Medical Imaging Analysis

Artificial General Intelligence (AGI) models, such as large language models (LLMs) exemplified by GPT-4, have achieved unprecedented success in general domain tasks. However, when these models are directly applied to specialized fields like medical imaging, they often encounter challenges arising from the inherent complexity and uniqueness of the medical domain. This review comprehensively explores the potential applications of AGI models in medical imaging and healthcare, focusing primarily on language, vision, and multimodal large models. The work provides a thorough overview of the key features and technologies underlying LLMs and AGI, and further examines the roadmap for the evolution and deployment of AGI models in healthcare. It summarizes their current applications, future potential, and associated challenges. Additionally, this review highlights promising future research directions, offering a comprehensive outlook on the prospects for AGI applications in medical imaging, healthcare, and related areas, and aims to provide insights into their broader implications for the future of these fields.

Figure 1. Timeline of Key Developments in AGI Technologies and Their Relationship to Medical Imaging

The related research findings, titled “Artificial General Intelligence for Medical Imaging Analysis,” have been published in IEEE Reviews in Biomedical Engineering. Prof. Xiang Li from Massachusetts General Hospital & Harvard Medical School, Dr. Lin Zhao from the research group of Prof. Tianming Liu at the University of Georgia, and Prof. Lu Zhang from Indiana University are the co-first authors. The corresponding authors are Prof. Xiang Li from Massachusetts General Hospital & Harvard Medical School, Prof. Tianming Liu from the University of Georgia, and Tenured Prof. Dinggang Shen from ShanghaiTech University. The primary affiliations include Massachusetts General Hospital, Harvard Medical School, the University of Georgia, and ShanghaiTech University. 

Paper link: https://arxiv.org/abs/2306.05480