大型语言模型在急性冠状动脉综合征指南上的表现,使用检索增强代
Michaella Alexandrou1, Sant Kumar2, Arun Umesh Mahtani3
1Minneapolis Heart Institute and Minneapolis Heart Institute Foundation, Abbott Northwestern Hospital, Minneapolis, Minnesota, USA.
JACC. Cardiovascular interventions
|October 29, 2025
概括
检索增强生成 (RAG) 在心脏病指南中显著提高了大语言模型 (LLM) 的准确性. 使用RAG的DeepSeek R1实现了94.7%的准确性,提高了临床决策潜力.
科学领域:
- 人工智能在医学中的应用
- 临床决策支持系统 临床决策支持系统
- 医疗信息学 医疗信息学
背景情况:
- 大型语言模型 (LLM) 在干预心脏病学中表现有前途,但受到事实上的不准确性 (幻觉) 的限制.
- 确保临床实用性需要提高医疗保健环境中LLM输出的可靠性和准确性.
研究的目的:
- 根据急性冠状动脉综合征指南,评估取回增强生成 (RAG) 对LLM回答问题的准确性的影响.
- 为了比较不同LLM的性能,有或没有RAG,与既定的临床指南.
主要方法:
- 三个LLM (ChatGPT-4o,DeepSeek R1,Med-PaLM 2) 用38个基于指南的心脏病问题进行评估.
- 聊天GPT-4o和DeepSeek R1在RAG和没有RAG的情况下进行了测试;Med-PaLM 2在没有RAG的情况下进行了测试.
- 模型响应与使用人工智能驱动的相似度评分工具对准则建议进行了比较.
主要成果:
- 使用RAG的DeepSeek R1以94.7%的准确率显示出最高的准确率,其次是使用RAG的ChatGPT-4o以92.1%的准确率.
- RAG显著提高了ChatGPT-4o的精度,从71.1%提高到92.1% (P = 0.017).
- 在没有RAG的情况下,DeepSeek R1 (78.9%) 的表现优于ChatGPT-4o (71.1%) 和Med-PaLM 2 (68.4%).
结论:
- 使用RAG将指南内容集成到LLM工作流中,提高了临床应用的准确性,特别是在干预心脏病学中.
- 通过RAG增强特定领域的知识,支持优化临床决策和遵守医疗指南.
- 当LLM与RAG增强时,它具有改善医疗保健实践和患者结果的巨大潜力.
相关概念视频
Acute Coronary Syndrome III: Diagnostic Studies
207
Diagnosing acute coronary syndrome or ACS begins with a thorough patient history. Notable symptoms include central, crushing chest pain radiating to the left arm, neck, jaw, or back, along with shortness of breath, sweating (diaphoresis), nausea, vomiting, dizziness, and palpitations.It is crucial to note any history of cardiac illnesses and assess risk factors, including age, gender, smoking, hypertension, diabetes, hyperlipidemia, and a sedentary lifestyle.During physical examination, vital...
207
Acute Coronary Syndrome I: Introduction
785
Acute Coronary Syndrome (ACS) encompasses a spectrum of heart conditions caused by sudden obstruction of coronary arteries, typically resulting from the rupture of an atherosclerotic plaque and subsequent thrombus (blood clot) formation. This obstruction can lead to partial or complete blockage of blood flow, causing varying degrees of myocardial ischemia or infarction.ACS includes the following clinical entities:Unstable Angina (UA)Non-ST-Elevation Myocardial Infarction (NSTEMI)ST-Elevation...
785
Improving Translational Accuracy
14.1K
Base complementarity between the three base pairs of mRNA codon and the tRNA anticodon is not a failsafe mechanism. Inaccuracies can range from a single mismatch to no correct base pairing at all. The free energy difference between the correct and nearly correct base pairs can be as small as 3 kcal/ mol. With complementarity being the only proofreading step, the estimated error frequency would be one wrong amino acid in every 100 amino acids incorporated. However, error frequencies observed in...
14.1K
Improving Translational Accuracy
3.5K
3.5K


