软件部署
急诊分诊台
计算机科学
构造(python库)
医疗保健
任务(项目管理)
数据科学
人工智能
大数据
知识管理
健康数据
生成语法
计算机安全
自然语言生成
对抗制
电子健康
医疗服务
任务分析
钥匙(锁)
作者
Qianye Lin,Yanzhi Song
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2025-08-01
卷期号:: 187-193
标识
DOI:10.1109/cme67420.2025.11239372
摘要
Intelligent medical systems improve healthcare efficiency and reduce costs by leveraging health data. Recently, large language models (LLMs) have significantly advanced intelligent medical services. However, challenges persist, including insufficient data quality, limited dataset scale, privacy concerns, and uncontrolled generative dialogue. To address these issues, we propose a knowledge-graph-based method for multi-turn dialogue generation using LLMs and construct the CMedConsul dataset, featuring over $\mathbf{2 0, 0 0 0}$ triage dialogues and $\mathbf{1 7, 0 0 0}$ pre-consultation dialogues to mitigate data scarcity. We then fine-tune LLMs on CMedConsul to develop an intelligent medical assistant CMedGuider, which integrates triage, pre-consultation, and EMR generation into a unified workflow. Experimental results show that it achieves $\mathbf{9 3. 2 \%}$ accuracy in the triage task and $\mathbf{9 5. 3 4 \%}$ accuracy in the EMR generation task, demonstrating its effectiveness and deployment potential. We open-source the datasets and model for future research.
科研通智能强力驱动
Strongly Powered by AbleSci AI