| 摘要: |
| ［摘要］ 2025年欧洲肿瘤内科学会（ESMO）发布的《ESMO人工智能大语言模型在肿瘤临床实践中的应用指南（ELCAP）》为大语言模型（LLMs）在肿瘤领域的规范化应用提供了首个国际化共识框架。该文旨在从医工融合专家视角深度解读其核心要旨，并探讨其在中国肿瘤临床实践中应用所面临的独特挑战。ELCAP通过德尔菲共识法，将LLMs应用划分为面向患者（type 1）、面向医护人员（type 2）和后台管理（type 3）三大体系，并达成了22条关键共识语句。笔者认为，该指南的发布标志着肿瘤人工智能（AI）从“自发探索”迈向“规范治理”，其核心价值在于确立了“人机协作、人类主导”的风险防控原则。针对中国临床环境，该文建议重点突破中文医疗大模型的垂直领域优化、建立符合本土伦理要求的分级监管体系及中医特色辅助决策接口，以推动生成式AI在精准肿瘤学中的可靠转化。 |
| 关键词: 人工智能 大语言模型 肿瘤学 临床实践 指南解读 |
| DOI：10.3969/j.issn.1674-3806.2026.01.02 |
| 分类号:R 73；TP 18 |
| 基金项目:深圳市科技计划项目（编号：JCYJ20220530153612028）；中国医学科学院肿瘤医院深圳医院科研课题（编号：E010222007） |
|
| Interpretation of the 2025 ESMO guidance on the use of Large Language Models in Clinical Practice(ELCAP) |
|
LIU Hu1, WU Bin2, NIU Niu3
|
|
1.Department of Medical Oncology, the Second Affiliated Hospital of Southern University of Science and Technology(the Third People′s Hospital of Shenzhen), Shenzhen 518112, China; 2.Department of Pulmonary and Critical Care Medicine, South China Hospital of Shenzhen University, Shenzhen 518110, China; 3.Department of Medical Oncology, Cancer Hospital Chinese Academy of Medical Sciences, Shenzhen, National Cancer Center, National Clinical Research Center for Cancer, Shenzhen 518116, China
|
| Abstract: |
| ［Abstract］ European Society for Medical Oncology(ESMO) guidance on the use of Large Language Models in Clinical Practice(ELCAP), released by ESMO in 2025(the 2025 ELCAP), provides the first international consensus framework for the standardized application of large language models(LLMs) in oncology. This paper aims to deeply interpret the core essence of the 2025 ELCAP from the perspective of experts in the integration of medicine and engineering, and explores the unique challenges faced by the 2025 ELCAP in the clinical application of oncology in China. The 2025 ELCAP utilizes a Delphi consensus process to categorize LLMs applications into three primary systems: patient-facing(type 1), healthcare professional-facing(type 2), and background management(type 3), and establishes 22 key consensus statements. The authors believe that the release of the 2025 ELCAP marks the transition of AI oncology from “spontaneous exploration” toward “standardized governance”. Its core value lies in the establishment of risk prevention and control principles based on “human-machine collaboration and human oversight”. In light of the clinical environment in China, this paper suggests focusing on the vertical optimization of Chinese medical LLMs, the establishment of hierarchical regulatory systems that align with local ethical requirements, and the development of auxiliary decision-making interfaces featuring Traditional Chinese Medicine(TCM) characteristics. These efforts aim to facilitate the reliable translation of generative AI within precision oncology. |
| Key words: Artificial intelligence(AI) Large language models(LLMs) Oncology Clinical practice Interpretation of guideline |
