European guideline informed RAG-based GPT-4 decision support tool in tumor board meetings for breast cancer treatment

IF 2.9 2区医学 Q2 ONCOLOGY

Ejso Pub Date : 2025-08-17 DOI:10.1016/j.ejso.2025.110384

Nuran Abdullayev , Jonathan Kottlors , Hikmat Habibov , Fahrettin Yilmaz , Chantal Zimmer , Nils Große Hokamp , Simon Lennartz , Vilayat Valiyev , Cavid Abbasli , Lukas Goertz , Jan Borggrefe , David Maintz , Koray Ersahin , Sebastian Sanduleanu

{"title":"European guideline informed RAG-based GPT-4 decision support tool in tumor board meetings for breast cancer treatment","authors":"Nuran Abdullayev , Jonathan Kottlors , Hikmat Habibov , Fahrettin Yilmaz , Chantal Zimmer , Nils Große Hokamp , Simon Lennartz , Vilayat Valiyev , Cavid Abbasli , Lukas Goertz , Jan Borggrefe , David Maintz , Koray Ersahin , Sebastian Sanduleanu","doi":"10.1016/j.ejso.2025.110384","DOIUrl":null,"url":null,"abstract":"<div><div>Breast cancer is the most common cancer among women worldwide. Treatment decision-making in multidisciplinary tumor boards is complex, involving integration of clinical guidelines, patient data, and preferences.</div><div>We retrospectively evaluated MammaBoardGPT, an few-shot in-context learning and Retrieval Augmented Generation (RAG) enhanced GPT-4 model with European guidelines and five Tumor Board labeled cases, against standard GPT-4 in 25 breast cancer cases discussed at a German hospital.</div><div>After recursive prompting, MammaBoardGPT achieved complete agreement with tumor board decisions in 84 % of cases, partial agreement in 16 % and no disagreements; standard GPT-4 showed 76 % complete agreement, 20 % partial agreement, and 4 % complete disagreements.</div><div>A significant difference was observed in MammaBoardGPT's (Stuart-Maxwell P = 0.0048) but not for GPT-4 agreement with tumor board decisions before versus after recursive prompting (Stuart-Maxwell P = 0.135), neither for MammaBoardGPT and GPT-4 in their tumor board decision agreement post-recursive prompting (Stuart-Maxwell P = 0.37).</div><div>These findings support MammaBoardGPT's potential for tumor board decision support, warranting prospective validation and real-time assessment.</div></div>","PeriodicalId":11522,"journal":{"name":"Ejso","volume":"51 11","pages":"Article 110384"},"PeriodicalIF":2.9000,"publicationDate":"2025-08-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Ejso","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0748798325008121","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ONCOLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Breast cancer is the most common cancer among women worldwide. Treatment decision-making in multidisciplinary tumor boards is complex, involving integration of clinical guidelines, patient data, and preferences.

We retrospectively evaluated MammaBoardGPT, an few-shot in-context learning and Retrieval Augmented Generation (RAG) enhanced GPT-4 model with European guidelines and five Tumor Board labeled cases, against standard GPT-4 in 25 breast cancer cases discussed at a German hospital.

After recursive prompting, MammaBoardGPT achieved complete agreement with tumor board decisions in 84 % of cases, partial agreement in 16 % and no disagreements; standard GPT-4 showed 76 % complete agreement, 20 % partial agreement, and 4 % complete disagreements.

A significant difference was observed in MammaBoardGPT's (Stuart-Maxwell P = 0.0048) but not for GPT-4 agreement with tumor board decisions before versus after recursive prompting (Stuart-Maxwell P = 0.135), neither for MammaBoardGPT and GPT-4 in their tumor board decision agreement post-recursive prompting (Stuart-Maxwell P = 0.37).

These findings support MammaBoardGPT's potential for tumor board decision support, warranting prospective validation and real-time assessment.

查看原文本刊更多论文

欧洲指南告知肿瘤委员会会议中基于rag的GPT-4决策支持工具用于乳腺癌治疗

乳腺癌是全世界女性中最常见的癌症。多学科肿瘤委员会的治疗决策是复杂的，涉及临床指南、患者数据和偏好的整合。我们回顾性地评估了mamaboardgpt，这是一种基于欧洲指南和5例肿瘤委员会标记病例的少量情境学习和检索增强生成（RAG）增强GPT-4模型，与德国一家医院讨论的25例乳腺癌病例的标准GPT-4相比。在递归提示后，mamaboardgpt在84%的病例中与肿瘤委员会的决定完全一致，16%的病例部分一致，没有分歧；标准GPT-4显示76%完全一致，20%部分一致，4%完全不一致。在mamaboardgpt中观察到显著差异（stewart - maxwell P = 0.0048），但在递归提示前后，GPT-4与肿瘤委员会决定的一致性没有显著差异（stewart - maxwell P = 0.135），在递归提示后，mamaboardgpt和GPT-4的肿瘤委员会决定一致性也没有显著差异（stewart - maxwell P = 0.37）。这些发现支持了mamaboardgpt作为肿瘤委员会决策支持的潜力，保证了前瞻性验证和实时评估。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Ejso 医学-外科

CiteScore

6.40

自引率

2.60%

发文量

1148

审稿时长

41 days

期刊介绍： JSO - European Journal of Surgical Oncology ("the Journal of Cancer Surgery") is the Official Journal of the European Society of Surgical Oncology and BASO ~ the Association for Cancer Surgery. The EJSO aims to advance surgical oncology research and practice through the publication of original research articles, review articles, editorials, debates and correspondence.