在急性胆囊炎的诊断中使用大型语言模型:评估准确性和遵守指南
Marta Goglia1,2, Arianna Cicolani1, Francesco Maria Carrano1
1Unit of General Surgery, Department of Surgical Sciences and Translational Medicine, Sapienza University of Rome, Rome, Italy.
The American surgeon
|March 12, 2025
概括
大型语言模型 (LLM) 在回答急性胆囊炎的临床问题方面表现有前途. 聊天GPT 4.0和双子座高级提供准确的信息,帮助医生教育和患者护理.
科学领域:
- 医疗信息学 医疗信息学
- 医疗保健中的人工智能
- 临床决策支持 临床决策支持
背景情况:
- 大型语言模型 (LLM) 是先进的AI工具,具有文本理解和生成的能力.
- 2018年东京指南 (TG18) 为诊断和治疗急性胆囊炎提供了一个框架.
- 根据既定的临床指南评估LLM准确性对于评估其实用性至关重要.
研究的目的:
- 评估商业大型语言模型 (LLM) 在响应急性胆囊炎临床查询时的准确性.
- 为了比较ChatGPT 4.0与Gemini Advanced和GPT-01-preview等较新的LLM的性能.
- 评估LLM与2018年东京指南的诊断和管理建议的一致性.
主要方法:
- 对ChatGPT 4.0,Gemini Advanced和GPT-01-预览提出了关于急性胆囊炎的十个临床问题.
- 八个问题是从2018年东京指南中得出的;两个是作者生成的.
- 两个作者独立对LLM的答案进行了四分级的准确性和全面性评分,其中三分之一的作者解决了差异.
主要成果:
- 聊天GPT 4.0为90%的问题提供了一致的答案,其中40%被评为"准确和全面",50%被评为"准确但不全面".
- 双子座高级显示了可比的准确性,没有一个模型产生"完全不准确"的答案.
- 两种模式的响应中有很大一部分被评为"部分准确,部分不准确",突出了需要改进的领域.
结论:
- 像ChatGPT 4.0和Gemini Advanced这样的LLM证明了准确解决急性胆囊炎相关的临床问题的潜力.
- 需要仔细实施和不断完善LLM,以利用其能力.
- 临床医学士可以成为医生教育和患者信息的宝贵资源,并有可能增强临床决策.
相关概念视频
01:29Chronic Pancreatitis II: Collaborative Care
54
The management of chronic pancreatitis is multifaceted, involving a comprehensive approach that includes thorough assessment, diagnostic testing, and a variety of management strategies.
Assessment:
Assessment:
54
01:28Peptic Ulcer Disease III: Clinical Manifestations and Diagnostic Studies
74
Peptic ulcer disease (PUD) presents with diverse symptoms depending on the location and severity of the ulcer. Clinical manifestations of peptic ulcer include dull pain and a burning sensation in the mid-epigastric region.
Few clinical manifestations differentiate gastric ulcers from duodenal ulcers. Distinctions in the location, timing, and pain relief are crucial for healthcare providers in differentiating between gastric and duodenal ulcers during clinical assessments.
Few clinical manifestations differentiate gastric ulcers from duodenal ulcers. Distinctions in the location, timing, and pain relief are crucial for healthcare providers in differentiating between gastric and duodenal ulcers during clinical assessments.
74