Diagnostic accuracy of vision-language models on Japanese diagnostic radiology, nuclear medicine, and interventional radiology specialty board examinations

Author:

Oura Tatsushi,Tatekawa HiroyukiORCID,Horiuchi Daisuke,Matsushita Shu,Takita Hirotaka,Atsukawa Natsuko,Mitsuyama Yasuhito,Yoshida Atsushi,Murai Kazuki,Tanaka Rikako,Shimono Taro,Yamamoto Akira,Miki Yukio,Ueda Daiju

Abstract

Abstract Purpose The performance of vision-language models (VLMs) with image interpretation capabilities, such as GPT-4 omni (GPT-4o), GPT-4 vision (GPT-4V), and Claude-3, has not been compared and remains unexplored in specialized radiological fields, including nuclear medicine and interventional radiology. This study aimed to evaluate and compare the diagnostic accuracy of various VLMs, including GPT-4 + GPT-4V, GPT-4o, Claude-3 Sonnet, and Claude-3 Opus, using Japanese diagnostic radiology, nuclear medicine, and interventional radiology (JDR, JNM, and JIR, respectively) board certification tests. Materials and methods In total, 383 questions from the JDR test (358 images), 300 from the JNM test (92 images), and 322 from the JIR test (96 images) from 2019 to 2023 were consecutively collected. The accuracy rates of the GPT-4 + GPT-4V, GPT-4o, Claude-3 Sonnet, and Claude-3 Opus were calculated for all questions or questions with images. The accuracy rates of the VLMs were compared using McNemar’s test. Results GPT-4o demonstrated the highest accuracy rates across all evaluations with the JDR (all questions, 49%; questions with images, 48%), JNM (all questions, 64%; questions with images, 59%), and JIR tests (all questions, 43%; questions with images, 34%), followed by Claude-3 Opus with the JDR (all questions, 40%; questions with images, 38%), JNM (all questions, 42%; questions with images, 43%), and JIR tests (all questions, 40%; questions with images, 30%). For all questions, McNemar’s test showed that GPT-4o significantly outperformed the other VLMs (all P < 0.007), except for Claude-3 Opus in the JIR test. For questions with images, GPT-4o outperformed the other VLMs in the JDR and JNM tests (all P < 0.001), except Claude-3 Opus in the JNM test. Conclusion The GPT-4o had the highest success rates for questions with images and all questions from the JDR, JNM, and JIR board certification tests.

Funder

Guerbet

Iida Group Holdings

Publisher

Springer Science and Business Media LLC

Reference24 articles.

1. Brown TB, Mann B, Ryder N, Subbiah M, Kaplan J, Dhariwal P, et al. Language Models are Few-Shot Learners [Internet]. arXiv [cs.CL]. 2020. Available from: http://arxiv.org/abs/2005.14165.

2. OpenAI. GPT-4 Technical Report [Internet]. arXiv [cs.CL]. 2023. Available from: http://arxiv.org/abs/2303.08774.

3. Eloundou T, Manning S, Mishkin P, Rock D. GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models [Internet]. arXiv [econ.GN]. 2023. Available from: http://arxiv.org/abs/2303.10130.

4. Anthropic AI. The claude 3 model family: Opus, sonnet, haiku. Claude-3 Model Card. 2024.

5. Takita H, Kabata D, Walston SL, Tatekawa H, Saito K, Tsujimoto Y, et al. Diagnostic performance comparison between generative AI and physicians: a systematic review and meta-analysis [Internet]. bioRxiv. 2024. Available from: https://doi.org/10.1101/2024.01.20.24301563v2.

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3