基于大语言模型的实验室报告解读系统构建研究  

Study on the Construction of a Laboratory Report Interpretation System Based on Large Language Model

作　　者：陆小琴 伍柯[2] 雷玉倩 唐胡 武宇翔 武永康 王莉[1,5] LU Xiaoqin;WU Ke;LEI Yuqian;TANG Hu;WU Yuxiang;WU Yongkang;WANG Li(Department of Laboratory Medicine,West China Hospital,Sichuan University,Chengdu 610041,China;Department of Information and Internet Hospital Management,Jintang First People’s Hospital,Chengdu 610400,China;School of Pharmacy and Laboratory Medicine,Ya’an Polytechnic College,Ya’an 625000,China;UWE College of Hainan Medical University,Haikou 571199,China;Department of Gynecology,West China Tianfu Hospital,Sichuan University,Chengdu 610213,China)

机构地区：[1]四川大学华西医院实验医学科,成都610041 [2]金堂县第一人民医院信息与互联网医院管理部,成都610400 [3]雅安职业技术学院药学与检验学院,雅安625000 [4]海南医科大学西英格兰学院,海口571199 [5]四川大学华西天府医院妇科,成都610213

出　　处：《医学信息学杂志》2025年第3期73-78,共6页Journal of Medical Informatics

基　　金：四川省科技计划项目(项目编号:2024YFFKO131)。

摘　　要：目的/意义构建智能实验室报告解读系统,帮助患者了解自身健康状况。方法/过程收集四川省金堂县第一人民医院实验室检验报告并标注,以ERNIE-4.0-Turbo-8K为基础模型,用标注数据监督微调,引入检索增强生成机制优化模型性能。结果/结论监督微调后,模型在实验室报告解读任务中多项评估指标优于微调前;以项目编码为索引的检索增强生成机制,能精准检索匹配知识切片,提升解读准确性与可解释性。基于大语言模型优化的智能系统,在提升医疗服务质量和效率方面潜力巨大,但准确性、稳定性尚待深入评估。Purpose/Significance To build an intelligent laboratory report interpretation system to help patients understand their own health conditions.Method/Process The study collects and annotates the laboratory test reports from Jintang First People’s Hospital in Sichuan province.Using ERNIE-4.0-Turbo-8K as the base model,it conducts supervised fine-tuning with the annotated data,and introduces the retrieval-augmented generation mechanism to optimize the model performance.Result/Conclusion After supervised fine-tuning,the model outperforms the pre-fine-tuning in multiple evaluation indicators for the laboratory report interpretation task.The retrieval-augmented generation mechanism indexed by item codes can accurately retrieve and match knowledge slices,improving the accuracy and interpretability of the interpretation.The optimized intelligent system based on large language models has great potential in improving the quality and efficiency of medical services.However,its accuracy and stability still need in-depth evaluation.

关 键 词：大语言模型 实验室报告解读 监督微调 检索增强生成 

分 类 号：R-058[医药卫生]