文献速递乳腺乳腺X线

商业人工智能与放射科医师在大规模人群乳腺X线摄影和断层合成筛查队列中的阴性预测值与召回率比较

Commercial Artificial Intelligence Versus Radiologists: NPV and Recall Rate in Large Population-Based Digital Mammography and Tomosynthesis Screening Mammography Cohorts.
PubMed 文献 · AJR Am J Roentgenol · 2025-12-17

该回顾性单机构研究纳入观察性人群 Athena 乳腺健康网络中的数字乳腺X线摄影(DM)和数字乳腺断层合成(DBT)筛查队列,比较商业人工智能系统在不同阈值下与放射科医师的阴性预测值和召回率。结果显示,人工智能在两种阳性阈值下均获得高阴性预测值,但召回率高于放射科医师;将中危检查纳入阳性会检出更多癌症但显著增加召回率。

要点

  • 在 DM 队列中,AI 以高危为阳性时召回率 14.0%,以中危或高危为阳性时召回率 41.8%,均高于放射科医师的 7.2%。
  • 在 DBT 队列中,AI 以高危为阳性时召回率 12.1%,以中危或高危为阳性时召回率 31.9%,均高于放射科医师的 6.9%。
  • 两种检查下 AI 的阴性预测值均达 99.8%–99.9%,与放射科医师的 99.9% 接近。

核心结论

  • 在大型 DM 和 DBT 队列中,AI 在两种诊断阈值下均获得高阴性预测值,但召回率高于放射科医师。

研究设计与样本

  • 回顾性单机构研究,纳入观察性人群 Athena 乳腺健康网络中 2010 年 1 月至 2019 年 12 月的筛查检查,采用分层随机抽样;DM 队列 26,693 次检查(20,409 名女性,平均年龄 58.1 岁),DBT 队列 4,824 次检查(4,379 名女性,平均年龄 61.3 岁);放射科医师判读取自临床报告,商业 AI 系统将检查分为低危、中危或高危,筛查后 1 年内的乳腺癌诊断来自州癌症登记处。

诊断标准与关键数值

  • DM 队列:AI 将 58.2%、27.7%、14.0% 的检查分别归为低危、中危、高危;放射科医师敏感度 88.6%、特异度 93.3%、召回率 7.2%、NPV 99.9%;AI 以高危为阳性时分别为 74.4%、86.3%、14.0%、99.8%;AI 以中危或高危为阳性时分别为 94.0%、58.6%、41.8%、99.9%。DBT 队列:AI 将 68.1%、19.8%、12.1% 的检查分别归为低危、中危、高危;放射科医师敏感度 83.8%、特异度 93.7%、召回率 6.9%、NPV 99.9%;AI 以高危为阳性时分别为 78.4%、88.4%、12.1%、99.8%;AI 以中危或高危为阳性时分别为 89.2%、68.5%、31.9%、99.8%。

临床应用与阅片要点

  • 作者指出,这些发现支持 AI 帮助放射科医师提高工作流程效率的潜力,但需要策略来处理频繁的假阳性结果,尤其是中危类别。

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

BACKGROUND. By reliably classifying screening mammograms as negative, artificial intelligence (AI) could minimize radiologists' time spent reviewing high volumes of normal examinations and help prioritize examinations with high likelihood of malignancy. OBJECTIVE. The purpose of this study was to compare performance of AI, classifying examinations as positive at different thresholds, with that of radiologists, focusing on NPV and recall rates, in large population-based digital mammography (DM) and digital breast tomosynthesis (DBT) screening cohorts. METHODS. This retrospective single-institution study included women enrolled in the observational population-based Athena Breast Health Network. Stratified random sampling was used to identify cohorts of DM and DBT screening examinations performed from January 2010 through December 2019. Radiologists' interpretations were extracted from clinical reports. A commercial AI system classified examinations as low, intermediate, or elevated risk. Breast cancer diagnoses within 1 year after screening examinations were identified from a state cancer registry. AI and radiologist performance were compared. RESULTS. The DM cohort included 26,693 examinations in 20,409 women (mean age, 58.1 years). AI classified 58.2%, 27.7%, and 14.0% of examinations as low, intermediate, and elevated risk, respectively. Sensitivity, specificity, recall rate, and NPV for radiologists were 88.6%, 93.3%, 7.2%, and 99.9%; for AI defining positive results as elevated risk, 74.4%, 86.3%, 14.0%, and 99.8%; and for AI defining positive results as intermediate or elevated risk, 94.0%, 58.6%, 41.8%, and 99.9%. The DBT cohort included 4824 examinations in 4379 women (mean age, 61.3 years). AI classified 68.1%, 19.8%, and 12.1% of examinations as low, intermediate, and elevated risk, respectively. Sensitivity, specificity, recall rate, and NPV for radiologists were 83.8%, 93.7%, 6.9%, and 99.9%; for AI defining positive results as elevated risk, 78.4%, 88.4%, 12.1%, and 99.8%; and for AI defining positive results as intermediate or elevated risk, 89.2%, 68.5%, 31.9%, and 99.8%. CONCLUSION. In large DM and DBT cohorts, AI at either diagnostic threshold achieved high NPV but had higher recall rates than radiologists. Defining positive AI results to include intermediate-risk examinations, versus only elevated-risk examinations, detected additional cancers but yielded markedly increased recall rates. CLINICAL IMPACT. The findings support AI's potential to aid radiologists' workflow efficiency. However, strategies are needed to address frequent false-positive results, particularly in the intermediate-risk category.

来源:AJR Am J Roentgenol,查看原文。版权归原作者所有。

引用

  1. Chen IE, Joines MM, Capiro N, et al. Commercial Artificial Intelligence Versus Radiologists: NPV and Recall Rate in Large Population-Based Digital Mammography and Tomosynthesis Screening Mammography Cohorts. AJR Am J Roentgenol. 2025. PMID: 40899676. doi:10.2214/AJR.25.32889

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。