Savage Cody H, Kanhere Adway, Parekh Vishwa, Langlotz Curtis P, Joshi Anupam, Huang Heng, Doo Florence X
From the University of Maryland Medical Intelligent Imaging (UM2ii) Center, Department of Diagnostic Radiology and Nuclear Medicine, University of Maryland School of Medicine, 22 S Greene St, Baltimore, MD 21201 (C.H.S., A.K., V.P., F.X.D.); Departments of Radiology, Medicine, and Biomedical Data Science, Stanford University, Palo Alto, Calif (C.P.L.); Department of Computer Science and Electrical Engineering, College of Engineering and Information Technology, University of Maryland, Baltimore County, Baltimore, Md (A.J.); Department of Computer Science, University of Maryland, College Park, College Park, Md (H.H.); and University of Maryland Institute for Health Computing, University of Maryland, North Bethesda, Md (H.H., F.X.D.).
Radiology. 2025 Jan;314(1):e241073. doi: 10.1148/radiol.241073.
Integrating large language models (LLMs) into health care holds substantial potential to enhance clinical workflows and care delivery. However, LLMs also pose serious risks if integration is not thoughtfully executed, with complex challenges spanning accuracy, accessibility, privacy, and regulation. Proprietary commercial LLMs (eg, GPT-4 [OpenAI], Claude 3 Sonnet and Claude 3 Opus [Anthropic], Gemini [Google]) have received much attention from researchers in the medical domain, including radiology. Interestingly, open-source LLMs (eg, Llama 3 and LLaVA-Med) have received comparatively little attention. Yet, open-source LLMs hold several key advantages over proprietary LLMs for medical institutions, hospitals, and individual researchers. The wider adoption of open-source LLMs has been slower, perhaps in part due to the lack of familiarity, accessible computational infrastructure, and community-built tools to streamline their local implementation and customize them for specific use cases. Thus, this article provides a tutorial for the implementation of open-source LLMs in radiology, including examples of commonly used tools for text generation and techniques for troubleshooting issues with prompt engineering, retrieval-augmented generation, and fine-tuning. Implementation-ready code for each tool is provided at . In addition, this article compares the benefits and drawbacks of open-source and proprietary LLMs, discusses the differentiating characteristics of popular open-source LLMs, and highlights recent advancements that may affect their adoption.
将大语言模型（LLMs）整合到医疗保健领域具有提升临床工作流程和护理服务的巨大潜力。然而，如果整合执行不当，大语言模型也会带来严重风险，涉及准确性、可及性、隐私和监管等复杂挑战。专有的商业大语言模型（如GPT-4[OpenAI]、Claude 3 Sonnet和Claude 3 Opus[Anthropic]、Gemini[谷歌]）受到了包括放射学在内的医学领域研究人员的广泛关注。有趣的是，开源大语言模型（如Llama 3和LLaVA-Med）受到的关注相对较少。然而，对于医疗机构、医院和个体研究人员而言，开源大语言模型相对于专有大语言模型具有几个关键优势。开源大语言模型的广泛采用速度较慢，这可能部分归因于缺乏熟悉度、可及的计算基础设施以及社区构建的工具来简化其本地实施并针对特定用例进行定制。因此，本文提供了一个在放射学中实施开源大语言模型的教程，包括文本生成常用工具的示例以及解决提示工程、检索增强生成和微调问题的技术。每个工具的可直接用于实施的代码可在[具体网址]获取。此外，本文比较了开源和专有大语言模型的优缺点，讨论了流行开源大语言模型的区别特征，并强调了可能影响其采用的最新进展。
