乳腺X线筛查中错误AI提示对阅片者行为的影响
本研究采用回顾性多阅片者配对设计,评估10名英国乳腺筛查医师在有无AI辅助下解读60例乳腺X线检查的差异。结果显示,AI假阴性提示显著降低阅片者敏感度(39% vs 71%,P=.002),假阳性提示提高特异度(39% vs 21%,P=.004)。错误AI提示同时改变阅片者的视觉搜索行为,提示AI阈值需谨慎校准。
要点
- AI假阴性提示会显著降低阅片者对乳腺癌的检出敏感度,临床应用中需警惕。
- AI假阳性提示虽提高特异度但可能掩盖真实病灶,报告时应结合原始图像独立判断。
- AI提示数量越多阅片时间越长,需合理设置显示阈值以平衡效率与准确度。
核心结论
- 错误的AI建议影响乳腺X线解读的准确性及视觉搜索行为,其中假阴性AI建议的负面效应最为显著,AI阈值应据此进行校准。
研究设计与样本
- 回顾性多阅片者配对研究,于2024年9月至2025年2月间进行。
- 纳入10名英国国家医疗服务体系乳腺筛查项目阅片医师,评估60例双侧双体位乳腺X线筛查检查。
- 测试集包含真阳性26例、假阴性14例、假阳性14例、真阴性6例AI建议,以3年随访或组织病理学为参考标准。
- 第一轮无AI辅助阅片,6周后第二轮使用商用AI工具辅助,显示区域评分≥10分的提示(评分范围0–100)。
- 采用眼动追踪设备记录阅片者在特定图像区域的注视行为。
关键影像表现/征象
- 阅片者在AI假阴性提示的病例中,对癌灶区域的注视频率较无辅助时降低(每秒注视次数0.44 vs 0.47,P=.03)。
- AI假阳性提示病例中,阅片者注视持续时间较无辅助时缩短(0.54秒 vs 0.56秒,P=.001)。
- 可见AI提示(真阳性与假阳性)数量越多,中位阅片时间越长,从无提示的25秒增至4个及以上提示的34秒(P=.001)。
诊断标准与关键数值
- AI假阴性提示时,中位阅片者敏感度为39%,低于无辅助阅片的71%(P=.002)。
- AI假阳性提示时,中位阅片者特异度为39%,高于无辅助阅片的21%(P=.004)。
- AI提示显示标准为区域评分≥10分(0–100量表)。
临床应用与阅片要点
- 报告时应意识到AI假阴性提示会引导视觉搜索偏离癌灶区域,需要对AI未标记区域保持主动搜寻习惯。
- AI假阳性提示缩短注视时间可能造成诊断满足,应始终保持系统性的双体位全面观察。
- 过多AI提示延长阅片时间,可通过提高显示阈值或过滤低置信度提示来优化工作流。
局限性
- 研究为回顾性设计,测试集仅60例,病例数有限。
- 所有阅片者来自英国国家医疗服务体系乳腺筛查项目,结果在其他阅片环境中的推广性不确定。
- 仅评估一款商用AI工具,不同AI系统的影响可能存在差异。
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
Background Incorrect artificial intelligence (AI) suggestions can lead to automation bias; however, their impact on medical image interpretation is underresearched. Purpose To assess how incorrect AI suggestions influence the diagnostic accuracy, read times, and visual search behavior of readers interpreting screening mammograms. Materials and Methods In this retrospective multireader paired study conducted between September 2024 and February 2025, 10 National Health Service Breast Screening Programme mammography readers evaluated a test set of two-view mammography screening examinations. The test set included true-positive (TP), false-negative (FN), false-positive (FP), and true-negative (TN) AI suggestions, verified by 3 years of follow-up or histopathologic analysis. In round 1, readers interpreted cases without AI. In round 2, conducted 6 weeks later, a commercially available AI tool was used as decision support, displaying prompts with a region score of 10 or higher (scale, 0-100). Eye-tracking cameras recorded readers' fixations-maintained gaze-over specific image areas. Wilcoxon signed rank tests were used for paired comparisons between rounds, and Kruskal-Wallis tests compared cases with different AI outcomes. Results The test set (n = 60) included cases with 26 TP, 14 FN, 14 FP, and six TN AI suggestions. Median reader sensitivity was lower for cases with FN AI suggestions when reading cases with AI (39%) compared with unassisted reading (71%; P = .002). Reader specificity was higher for cases with FP AI suggestions (39% vs 21%; P = .004). A greater number of visible (TP and FP) AI prompts led to longer median read times, from 25 seconds (zero prompts) to 34 seconds (four or more prompts) (P = .001). Readers fixated less when reviewing cancer cases that AI failed to detect (FN suggestions) compared with unassisted reading (0.44 vs 0.47 fixations per second; P = .03). Shorter fixation durations were observed when readers interpreted cases with FP AI suggestions compared with unassisted reading (0.54 vs 0.56 second; P = .001). Conclusion Incorrect AI suggestions influenced both reader accuracy and visual search behaviors during mammography interpretation. The greatest negative impact was observed with FN AI suggestions; therefore, AI thresholds should be calibrated accordingly. © RSNA, 2026 Supplemental material is available for this article. See also the editorial by Clauser in this issue. See also the editorial by Abbasi and Giess in this issue.
引用
- Taib, Partridge, Phillips, et al. Automation Bias in Action: Eye Tracking of Humans Reading Screening Mammograms with and without AI Prompts. Radiology. 2026. PMID: 42446361. doi:10.1148/radiol.252590
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。