AtlasGPT: a language model grounded in neurosurgery with domain-specific data and document retrieval.

IF 3.5 2区 医学 Q1 CLINICAL NEUROLOGY
Rohaid Ali, Hael F Abdulrazeq, Advait Patil, Morgan Cheatham, Ian D Connolly, Oliver Y Tang, Cody A Doberstein, Tori Riccelli, Kevin T Huang, Ganesh M Shankar, Theresa Williamson, John H Shin, Bob Carter, Radmehr Torabi, Christine K Lee, Deus Cielo, Albert E Telfeian, Ziya L Gokaslan, Aaron A Cohen-Gadol, James Zou, Wael F Asaad
{"title":"AtlasGPT: a language model grounded in neurosurgery with domain-specific data and document retrieval.","authors":"Rohaid Ali, Hael F Abdulrazeq, Advait Patil, Morgan Cheatham, Ian D Connolly, Oliver Y Tang, Cody A Doberstein, Tori Riccelli, Kevin T Huang, Ganesh M Shankar, Theresa Williamson, John H Shin, Bob Carter, Radmehr Torabi, Christine K Lee, Deus Cielo, Albert E Telfeian, Ziya L Gokaslan, Aaron A Cohen-Gadol, James Zou, Wael F Asaad","doi":"10.3171/2024.12.JNS241607","DOIUrl":null,"url":null,"abstract":"<p><strong>Objective: </strong>Large language models (LLMs) have shown promising performance on medical licensing examinations, but their ability to excel in subspecialty domains and their robustness under adversarial conditions remain unclear. Herein, the authors present AtlasGPT, a subspecialty-focused LLM for neurosurgery, and evaluate its performance on a benchmark multiple-choice question bank and under adversarial testing, as well as its ability to generate high-quality explanations.</p><p><strong>Methods: </strong>AtlasGPT was built by fine-tuning GPT-4 architecture and retrieval-augmented generation from neurosurgical knowledge sources. Its performance was compared with that of GPT-4 and Gemini Advanced on a 149-question neurosurgery examination. Adversarial testing assessed robustness to misinformation. Answer explanations were rated by 15 independent neurosurgeons and compared with the question bank.</p><p><strong>Results: </strong>Across all 149 questions and on text-only questions, AtlasGPT (96%) outperformed Gemini Advanced (93%) and GPT-4 (88%) in accuracy. In adversarial testing, under which AtlasGPT was tasked with identifying medical misinformation, it was fooled 14% of the time, compared with 44% for GPT-4 and 68% for Gemini Advanced. Neurosurgeons rated AtlasGPT's answer explanations as significantly more comprehensive, relevant, and better referenced than the question bank's explanations of the responses (p < 0.001). AtlasGPT did not demonstrate any evidence of hallucination or other content that would be harmful for patient care or the surgeon's clinical decision.</p><p><strong>Conclusions: </strong>AtlasGPT demonstrates the potential of subspecialty-focused LLMs to outperform general models, exhibit robustness to misinformation, and generate high-quality explanations. Domain-specific LLMs may improve medical knowledge, decision-making, and educational materials in complex fields like neurosurgery.</p>","PeriodicalId":16505,"journal":{"name":"Journal of neurosurgery","volume":" ","pages":"1-8"},"PeriodicalIF":3.5000,"publicationDate":"2025-04-18","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of neurosurgery","FirstCategoryId":"3","ListUrlMain":"https://doi.org/10.3171/2024.12.JNS241607","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}
引用次数: 0

Abstract

Objective: Large language models (LLMs) have shown promising performance on medical licensing examinations, but their ability to excel in subspecialty domains and their robustness under adversarial conditions remain unclear. Herein, the authors present AtlasGPT, a subspecialty-focused LLM for neurosurgery, and evaluate its performance on a benchmark multiple-choice question bank and under adversarial testing, as well as its ability to generate high-quality explanations.

Methods: AtlasGPT was built by fine-tuning GPT-4 architecture and retrieval-augmented generation from neurosurgical knowledge sources. Its performance was compared with that of GPT-4 and Gemini Advanced on a 149-question neurosurgery examination. Adversarial testing assessed robustness to misinformation. Answer explanations were rated by 15 independent neurosurgeons and compared with the question bank.

Results: Across all 149 questions and on text-only questions, AtlasGPT (96%) outperformed Gemini Advanced (93%) and GPT-4 (88%) in accuracy. In adversarial testing, under which AtlasGPT was tasked with identifying medical misinformation, it was fooled 14% of the time, compared with 44% for GPT-4 and 68% for Gemini Advanced. Neurosurgeons rated AtlasGPT's answer explanations as significantly more comprehensive, relevant, and better referenced than the question bank's explanations of the responses (p < 0.001). AtlasGPT did not demonstrate any evidence of hallucination or other content that would be harmful for patient care or the surgeon's clinical decision.

Conclusions: AtlasGPT demonstrates the potential of subspecialty-focused LLMs to outperform general models, exhibit robustness to misinformation, and generate high-quality explanations. Domain-specific LLMs may improve medical knowledge, decision-making, and educational materials in complex fields like neurosurgery.

AtlasGPT:一种基于神经外科领域特定数据和文档检索的语言模型。
目的:大型语言模型(LLMs)在医学执照考试中表现良好,但它们在亚专业领域的表现能力及其在对抗条件下的鲁棒性尚不清楚。在此,作者介绍了一种专注于神经外科的亚专业法学硕士AtlasGPT,并评估了它在基准选择题库和对抗性测试中的表现,以及它生成高质量解释的能力。方法:通过对GPT-4结构进行微调和从神经外科知识库中检索增强生成,构建AtlasGPT。在一项149题的神经外科检查中,将其表现与GPT-4和Gemini Advanced进行了比较。对抗性测试评估了对错误信息的稳健性。答案解释由15名独立神经外科医生评分,并与题库进行比较。结果:在所有149个问题和纯文本问题中,AtlasGPT(96%)在准确率上优于Gemini Advanced(93%)和GPT-4(88%)。在对抗性测试中,AtlasGPT的任务是识别医疗错误信息,它被欺骗的几率为14%,而GPT-4和Gemini Advanced分别为44%和68%。神经外科医生认为AtlasGPT的答案解释比题库对答案的解释更全面、更相关、更有参考价值(p < 0.001)。AtlasGPT没有显示出任何幻觉或其他有害于患者护理或外科医生临床决定的内容的证据。结论:AtlasGPT证明了以亚专业为重点的法学硕士的潜力,其表现优于一般模型,对错误信息表现出鲁棒性,并产生高质量的解释。特定领域的法学硕士可以提高医学知识,决策和复杂领域的教育材料,如神经外科。
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
Journal of neurosurgery
Journal of neurosurgery 医学-临床神经学
CiteScore
7.20
自引率
7.30%
发文量
1003
审稿时长
1 months
期刊介绍: The Journal of Neurosurgery, Journal of Neurosurgery: Spine, Journal of Neurosurgery: Pediatrics, and Neurosurgical Focus are devoted to the publication of original works relating primarily to neurosurgery, including studies in clinical neurophysiology, organic neurology, ophthalmology, radiology, pathology, and molecular biology. The Editors and Editorial Boards encourage submission of clinical and laboratory studies. Other manuscripts accepted for review include technical notes on instruments or equipment that are innovative or useful to clinicians and researchers in the field of neuroscience; papers describing unusual cases; manuscripts on historical persons or events related to neurosurgery; and in Neurosurgical Focus, occasional reviews. Letters to the Editor commenting on articles recently published in the Journal of Neurosurgery, Journal of Neurosurgery: Spine, and Journal of Neurosurgery: Pediatrics are welcome.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信