评估ChatGPT在放射肿瘤学临床决策中的表现:开放式问题与图像
本研究前瞻性构建164个开放式临床问题(涵盖指南知识、复杂临床场景、护理、技术领域)并提交给ChatGPT-4和4O,按1-4级评分;另将20例鼻咽癌患者的MR图像提交给ChatGPT-4O进行T分期。结果显示,实用(1级)答案比例在护理和指南知识领域较高,在复杂临床场景和技术领域较低;ChatGPT-4O未能对鼻咽癌MR图像进行T分期,将所有实际非T4病例均判为T4(κ=0;95%CI,-0.253至0.253)。
要点
- ChatGPT-4和4O在护理和指南知识领域实用答案比例较高,复杂临床场景和技术领域较低。
- ChatGPT-4O未能对鼻咽癌MR图像进行T分期,将所有非T4病例判为T4(κ=0)。
核心结论
- ChatGPT可作为放射肿瘤学安全的临床决策支持工具,因其正确回答了绝大多数临床问题。
- 但在复杂临床场景和技术领域,其临床实用性应谨慎权衡。
- ChatGPT-4O尚不成熟,不能用于诊断图像进行癌症分期。
研究设计与样本
- 前瞻性构建164个开放式问题,涵盖指南知识、复杂临床场景、护理、技术领域,由专家制定并提交给ChatGPT-4和4O。
- 答案按1级(可直接用于临床决策)、2级(正确但不充分)、3级(正确与错误信息混合)、4级(完全错误)评分。
- 将20例不同T分期鼻咽癌患者的代表性诊断MR图像提交给ChatGPT-4O,要求确定每例的T分期。
主要发现
- 护理领域实用答案比例:GPT-4为91.9%,GPT-4O为94.6%;指南知识领域:GPT-4为82.2%,GPT-4O为88.9%。
- 复杂临床场景领域实用答案比例:GPT-4为54.1%,GPT-4O为62.2%;技术领域:GPT-4为64.4%,GPT-4O为77.8%。
- ChatGPT-4O未能对鼻咽癌MR图像进行T分期,将所有实际非T4病例均判为T4(κ=0;95%CI,-0.253至0.253)。
临床怎么用
- ChatGPT可作为放射肿瘤学临床决策支持工具,因其正确回答了绝大多数临床问题。
- 在复杂临床场景和技术领域,应谨慎权衡ChatGPT的临床实用性。
- ChatGPT-4O尚不成熟,不能用于诊断图像进行癌症分期。
局限性
- 原文未述局限性
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)。
原文
PURPOSE: This study assesses the practicality and correctness of Chat Generative Pre-trained Transformer (ChatGPT)-4 and 4O's answers to clinical inquiries in radiation oncology, and evaluates ChatGPT-4O for staging nasopharyngeal carcinoma (NPC) cases with magnetic resonance (MR) images. METHODS AND MATERIALS: A total of 164 open-ended questions covering representative professional domains (Clinical_G: knowledge on standardized guidelines; Clinical_C: complex clinical scenarios; Nursing: nursing and health education; and Technology: radiation technology and dosimetry) were prospectively formulated by experts and presented to ChatGPT-4 and 4O. Each ChatGPT's answer was graded as 1 (Directly practical for clinical decision-making), 2 (Correct but inadequate), 3 (Mixed with correct and incorrect information), or 4 (Completely incorrect). ChatGPT-4O was presented with the representative diagnostic MR images of 20 patients with NPC across different T stages, and asked to determine the T stage of each case. RESULTS: The proportions of ChatGPT's answers that were practical (grade 1) varied across professional domains (P < .01), higher in Nursing (GPT-4: 91.9%; GPT-4O: 94.6%) and Clinical_G (GPT-4: 82.2%; GPT-4O: 88.9%) domains than in Clinical_C (GPT-4: 54.1%; GPT-4O: 62.2%) and Technology (GPT-4: 64.4%; GPT-4O: 77.8%) domains. The proportions of correct (grade 1+2) answers (GPT-4: 89.6%; GPT-4O: 98.8%; P < .01) were universally high across all professional domains. However, ChatGPT-4O failed to stage NPC cases via MR images, indiscriminately assigning T4 to all actually non-T4 cases (κ = 0; 95% CI, -0.253 to 0.253). CONCLUSIONS: ChatGPT could be a safe clinical decision-support tool in radiation oncology, because it correctly answered the vast majority of clinical inquiries across professional domains. However, its clinical practicality should be cautiously weighted particularly in the Clinical_C and Technology domains. ChatGPT-4O is not yet mature to interpret diagnostic images for cancer staging.
引用
- Chuang WK, Kao YS, Liu YT, et al. Assessing ChatGPT for Clinical Decision-Making in Radiation Oncology, With Open-Ended Questions and Images. Pract Radiat Oncol. 2025. PMID: 40311921. doi:10.1016/j.prro.2025.04.009
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。