GPT-4o与专用人工智能在乳腺超声成像中的比较研究:准确性、一致性、局限性与诊断潜力
本研究评估了ChatGPT(GPT-4o)与基于ChatGPT训练的乳腺超声助手(BUH)对232幅经组织病理学证实的乳腺良恶性病变超声图像的分析能力,由两名放射科医师共识评估。BUH在多个特征上表现更优且kappa值更高,但两者与放射科医师共识的一致性均有限,BI-RADS分类一致性GPT-4o略高(69.4%对65.9%),良恶性鉴别BUH略优(67.7%对65.9%)。作者认为两者与放射科医师评估及BI-RADS分类的兼容性有限,临床应用于乳腺超声解读仍为时过早且不可靠。
要点
- BUH在多个超声特征上与放射科医师共识的kappa值更高且差异有统计学意义(P .001)。
- 两者总体一致性相似:BUH κ 0.387-0.755,GPT-4o κ 0.317-0.803。
- BI-RADS分类一致性GPT-4o 69.4%略高于BUH 65.9%,良恶性鉴别BUH 67.7%略高于GPT-4o 65.9%。
核心结论
- 两种AI工具在超声图像分析中表现中等至良好,但与放射科医师评估及BI-RADS分类的兼容性有限,作者认为临床应用于乳腺超声解读仍为时过早且不可靠。
研究设计与样本
- 比较研究,纳入232幅经组织病理学证实的乳腺癌与纤维腺瘤患者超声图像,其中恶性133幅(57.3%)、良性99幅(42.7%)。
关键影像表现/征象
- 评估特征包括病变形状、方向、边缘、内部回声模式、回声强度、后方声学特征、微钙化或高回声灶、病灶周围高回声晕、水肿或结构扭曲、病变大小及BI-RADS分类。
- BUH在多个特征上与放射科医师共识的一致性kappa值更高且差异有统计学意义(P .001)。
- 两者总体一致性相似:BUH κ 0.387-0.755,GPT-4o κ 0.317-0.803。
诊断标准与关键数值
- BUH与放射科医师共识的BI-RADS分类一致率为65.9%,GPT-4o为69.4%。
- BUH良恶性鉴别成功率为67.7%,GPT-4o为65.9%。
- BUH各特征与共识的kappa值为0.387-0.755,GPT-4o为0.317-0.803。
临床应用与阅片要点
- 作者认为两者与放射科医师评估及BI-RADS分类的兼容性有限,临床应用于乳腺超声解读仍为时过早且不可靠。
局限性
- 作者指出两种AI工具与放射科医师评估及BI-RADS分类的兼容性有限。
- 作者认为其临床应用于乳腺超声解读仍为时过早且不可靠。
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: This study aimed to evaluate the ability of ChatGPT and Breast Ultrasound Helper, a special ChatGPT-based subprogram trained on ultrasound image analysis, to analyze and differentiate benign and malignant breast lesions on ultrasound images.
METHODS: Ultrasound images of histopathologically confirmed breast cancer and fibroadenoma patients were read GPT-4o (the latest ChatGPT version) and Breast Ultrasound Helper (BUH), a tool from the "Explore" section of ChatGPT. Both were prompted in English using ACR BI-RADS Breast Ultrasound Lexicon criteria: lesion shape, orientation, margin, internal echo pattern, echogenicity, posterior acoustic features, microcalcifications or hyperechoic foci, perilesional hyperechoic rim, edema or architectural distortion, lesion size, and BI-RADS category. Two experienced radiologists evaluated the images and the responses of the programs in consensus. The outputs, BI-RADS category agreement, and benign/malignant discrimination were statistically compared.
RESULTS: A total of 232 ultrasound images were analyzed, of which 133 (57.3%) were malignant and 99 (42.7%) benign. In comparative analysis, BUH showed superior performance overall, with higher kappa values and statistically significant results across multiple features (P .001). However, the overall level of agreement with the radiologists' consensus for all features was similar for BUH (κ: 0.387-0.755) and GPT-4o (κ: 0.317-0.803). On the other hand, BI-RADS category agreement was slightly higher in GPT-4o than in BUH (69.4% versus 65.9%), but BUH was slightly more successful in distinguishing benign lesions from malignant lesions (65.9% versus 67.7%).
CONCLUSIONS: Although both AI tools show moderate-good performance in ultrasound image analysis, their limited compatibility with radiologists' evaluations and BI-RADS categorization suggests that their clinical application in breast ultrasound interpretation is still early and unreliable.
引用
- Sanli DET, Sanli AN, Buyukdereli Atadag Y, et al. GPT-4o and Specialized AI in Breast Ultrasound Imaging: A Comparative Study on Accuracy, Agreement, Limitations, and Diagnostic Potential. J Ultrasound Med. 2025. PMID: 40548624. doi:10.1002/jum.16749
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。