ChatGPT-4 在 MRI 方案选择上与认证神经放射科医师高度一致
本研究回顾性纳入本院 1001 份、独立综合放射科 200 份及国外神经放射科 300 份真实放射申请单,评估 ChatGPT-4 选择 MRI 方案及节省时间方案的能力。结果显示方案建议几乎均被接受,读者间加权 kappa 为 0.88–0.98;自由组合方案在本院 766/1001(76.5%)和外部 300 份中 140/300(46.67%)获批准,平均每例节省 3:51 分钟和 2:59 分钟。
要点
- ChatGPT-4 基于真实申请单选择的 MRI 方案与认证神经放射科医师高度一致,加权 kappa 0.88–0.98。
- 本院 1001 份建议中仅 2 份(读者1)和 7 份(读者2)被评为不可接受,独立与外部数据集均无不可接受建议。
- 自由组合的节省时间方案在本院 76.5%、外部 46.67% 的病例中获批准,平均节省约 3–4 分钟。
核心结论
- ChatGPT-4 在选择 MRI 方案方面与认证(神经)放射科医师高度一致,并能从可用序列中提出获批准的节省时间方案。
研究设计与样本
- 回顾性研究,纳入本院神经放射科 1001 份、独立综合放射科 200 份和国外神经放射科 300 份真实放射申请单,提取年龄、性别和临床信息提示 ChatGPT-4 从预设机构列表中选择 MRI 方案,由四名独立评估者评分,并额外要求其制定节省时间的病例特异性方案。
诊断标准与关键数值
- 本院数据集 1001 份方案建议中,读者1 和读者2 分别有 2 份和 7 份被评为不可接受。
- 独立数据集和外部数据集均无方案建议被评为不可接受。
- 读者间一致性 Cohen 加权 kappa 范围为 0.88 至 0.98(各 p < 0.001)。
- ChatGPT-4 自由组合的方案在本院 766/1001(76.5%)和外部数据集 140/300(46.67%)的病例中获批准。
- 获采纳方案的平均节省时间(标准差)在本院为 3:51(分:秒)(±2:40)分钟,在外部为 2:59(±3:42)分钟。
临床应用与阅片要点
- ChatGPT-4 可根据真实申请单中的年龄、性别和临床信息从预设列表中选择 MRI 方案,与认证神经放射科医师高度一致。
- 其自由组合的节省时间方案在本院 76.5%、外部 46.67% 的病例中获批准,平均每例节省 3:51 分钟和 2:59 分钟。
局限性
- 原文未报告局限性。
以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。
英文原文(摘要)· 点击展开
OBJECTIVES: The aim of this study was to determine whether ChatGPT-4 can correctly suggest MRI protocols and additional MRI sequences based on real-world Radiology Request Forms (RRFs) as well as to investigate the ability of ChatGPT-4 to suggest time saving protocols.
MATERIAL & METHODS: Retrospectively, 1,001 RRFs of our Department of Neuroradiology (in-house dataset), 200 RRFs of an independent Department of General Radiology (independent dataset) and 300 RRFs from an external, foreign Department of Neuroradiology (external dataset) were included. Patients' age, sex, and clinical information were extracted from the RRFs and used to prompt ChatGPT- 4 to choose an adequate MRI protocol from predefined institutional lists. Four independent raters then assessed its performance. Additionally, ChatGPT-4 was tasked with creating case-specific protocols aimed at saving time.
RESULTS: Two and 7 of 1,001 protocol suggestions of ChatGPT-4 were rated "unacceptable" in the in-house dataset for reader 1 and 2, respectively. No protocol suggestions were rated "unacceptable" in both the independent and external dataset. When assessing the inter-reader agreement, Coheńs weighted ĸ ranged from 0.88 to 0.98 (each p < 0.001). ChatGPT-4's freely composed protocols were approved in 766/1,001 (76.5 %) and 140/300 (46.67 %) cases of the in-house and external dataset with mean time savings (standard deviation) of 3:51 (minutes:seconds) (±2:40) minutes and 2:59 (±3:42) minutes per adopted in-house and external MRI protocol.
CONCLUSION: ChatGPT-4 demonstrated a very high agreement with board-certified (neuro-)radiologists in selecting MRI protocols and was able to suggest approved time saving protocols from the set of available sequences.
引用
- Bendella Z, Wichtmann BD, Clauberg R, et al. Chat GPT-4 shows high agreement in MRI protocol selection compared to board-certified neuroradiologists. Eur J Radiol. 2025. PMID: 40961911. doi:10.1016/j.ejrad.2025.112416
供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。