人工智能在乳腺癌 X 光筛查中应用的早期影响指标
Early Indicators of the Impact of Using AI in Mammography Screening for Breast Cancer.
机构信息
From the Departments of Computer Science (A.D.L., M.L., M.N.) and Public Health (E.L.), University of Copenhagen, Copenhagen, Denmark; Department of Breast Examinations, Gentofte Hospital, Kildegårdsvej 30A, 2900 Hellerup, Denmark (A.D.L., I.V.); Department of Radiology and Nuclear Medicine, Radboud University Medical Center, Nijmegen, the Netherlands (N.K.); and ScreenPoint Medical, Nijmegen, the Netherlands (N.K.).
出版信息
Radiology. 2024 Jun;311(3):e232479. doi: 10.1148/radiol.232479.
Background Retrospective studies have suggested that using artificial intelligence (AI) may decrease the workload of radiologists while preserving mammography screening performance. Purpose To compare workload and screening performance for two cohorts of women who underwent screening before and after AI system implementation. Materials and Methods This retrospective study included 50-69-year-old women who underwent biennial mammography screening in the Capital Region of Denmark. Before AI system implementation (October 1, 2020, to November 17, 2021), all screenings involved double reading. For screenings conducted after AI system implementation (November 18, 2021, to October 17, 2022), likely normal screenings (AI examination score ≤5 before May 3, 2022, or ≤7 on or after May 3, 2022) were single read by one of 19 senior full-time breast radiologists. The remaining screenings were read by two radiologists with AI-assisted decision support. Biopsy and surgical outcomes were retrieved between October 1, 2020, and April 15, 2023, ensuring at least 180 days of follow-up. Screening metrics were compared using the χ test. Reading workload reduction was measured as saved screening reads. Results In total, 60 751 and 58 246 women were screened before and after AI system implementation, respectively (median age, 58 years [IQR, 54-64 years] for both cohorts), with a median screening interval before AI of 845 days (IQR, 820-878 days) and with AI of 993 days (IQR, 968-1013 days; < .001). After AI system implementation, the recall rate decreased by 20.5% (3.09% before AI [1875 of 60 751] vs 2.46% with AI [1430 of 58 246]; < .001), the cancer detection rate increased (0.70% [423 of 60 751] vs 0.82% [480 of 58 246]; = .01), the false-positive rate decreased (2.39% [1452 of 60 751] vs 1.63% [950 of 58 246]; < .001), the positive predictive value increased (22.6% [423 of 1875] vs 33.6% [480 of 1430]; < .001), the rate of small cancers (≤1 cm) increased (36.6% [127 of 347] vs 44.9% [164 of 365]; = .02), the rate of node-negative cancers was unchanged (76.7% [253 of 330] vs 77.8% [273 of 351]; = .73), and the rate of invasive cancers decreased (84.9% [359 of 423] vs 79.6% [382 of 480]; = .04). The reading workload was reduced by 33.5% (38 977 of 116 492 reads). Conclusion In a population-based mammography screening program, using AI reduced the overall workload of breast radiologists while improving screening performance. Published under a CC BY 4.0 license. See also the editorial by Lee and Friedewald in this issue.
背景 回顾性研究表明，使用人工智能（AI）可能会降低放射科医生的工作量，同时保持乳房 X 线摄影筛查的性能。
目的 比较两组女性在实施 AI 系统前后的工作量和筛查性能，这些女性接受了丹麦首都地区的两年一次的乳房 X 线摄影筛查。
材料和方法 本回顾性研究纳入了 50-69 岁的女性，她们在丹麦首都地区接受了两年一次的乳房 X 线摄影筛查。在 AI 系统实施之前（2020 年 10 月 1 日至 2021 年 11 月 17 日），所有筛查都需要进行双读。在 AI 系统实施之后（2021 年 11 月 18 日至 2022 年 10 月 17 日），对于可能正常的筛查（在 2022 年 5 月 3 日之前，AI 检查评分≤5，或在 2022 年 5 月 3 日或之后，AI 检查评分≤7），由 19 名全职的资深乳房放射科医生之一进行单读。其余的筛查由两名具有 AI 辅助决策支持的放射科医生进行阅读。在 2020 年 10 月 1 日至 2023 年 4 月 15 日期间，检索了活检和手术结果，确保了至少 180 天的随访。使用卡方检验比较了筛查指标。通过测量节省的筛查读片来衡量阅读工作量的减少。
结果 在 AI 系统实施之前和之后，分别有 60751 名和 58246 名女性接受了筛查（中位年龄，58 岁[IQR，54-64 岁]，两个队列均），在 AI 之前的中位筛查间隔为 845 天（IQR，820-878 天），在 AI 之后为 993 天（IQR，968-1013 天； <.001）。在 AI 系统实施之后，召回率下降了 20.5%（AI 前为 3.09%[1875 例/60751 例]，AI 后为 2.46%[1430 例/58246 例]； <.001），癌症检出率增加（0.70%[423 例/60751 例] vs 0.82%[480 例/58246 例]； =.01），假阳性率降低（2.39%[1452 例/60751 例] vs 1.63%[950 例/58246 例]； <.001），阳性预测值增加（22.6%[423 例/1875 例] vs 33.6%[480 例/1430 例]； <.001），小癌症（≤1cm）的检出率增加（36.6%[127 例/347 例] vs 44.9%[164 例/365 例]； =.02），淋巴结阴性癌症的检出率不变（76.7%[253 例/330 例] vs 77.8%[273 例/351 例]； =.73），浸润性癌症的检出率降低（84.9%[359 例/423 例] vs 79.6%[382 例/480 例]； =.04）。阅读工作量减少了 33.5%（38977 例/116492 例）。
结论 在基于人群的乳房 X 线摄影筛查项目中，使用 AI 降低了乳腺放射科医生的总体工作量，同时提高了筛查性能。
在知识共享署名 4.0 国际许可协议下发布。本研究中还刊登了 Lee 和 Friedewald 的社论。
Zotero 插件