A review of applying large language models in healthcare

Abstract

In response to the growing demand for healthcare and the increasing importance people place on medical services, efficiently meeting these needs within the constraints of limited healthcare resources is of great social and economic benefit. Therefore, research into applying Large Language Models (LLMs) in the healthcare sector holds significant importance. This paper provides a review of the research progress on the application of LLMs in the healthcare field. First, the basic framework of LLMs is summarized, and the training process of LLMs in healthcare is systematically reviewed. Next, six specific application areas of LLMs in healthcare are reviewed: disease diagnosis and decision support, dissemination of medical knowledge, medical assistance, medical image analysis, biomedicine, and medical education. Then, several representative healthcare-specific LLMs are discussed, along with their performance analysis. Following this, the challenges faced by LLMs in healthcare are summarized, and relevant suggestions are provided. The future development trends of LLMs in healthcare are also explored. Finally, a bibliometric analysis is performed. Through the literature review, we found: 1) After pretraining, LLMs are widely adaptable to downstream tasks, significantly enhancing processing performance and efficiency; 2) LLMs in healthcare possess multiple capabilities and can handle multimodal data; 3) Bibliometric analysis shows that researchers are paying increasing attention to the application of LLMs in healthcare; 4) Further research is needed in optimizing, improving reliability, and expanding practical applications of large healthcare models.