AbstractArtificial intelligence (AI) readers of mammograms compare favourably to individual radiologists in detecting breast cancer. However, AI readers cannot perform at the level of multi-reader systems used by screening programs in countries such as Australia, Sweden, and the UK. Therefore, implementation demands human-AI collaboration.

摘要乳腺X线照片的人工智能（AI）阅读器在检测乳腺癌方面优于个体放射科医生。然而，人工智能阅读器的性能无法达到澳大利亚、瑞典和英国等国家的筛选程序所使用的多阅读器系统的水平。因此，实施需要人工智能协作。

Here, we use a large, high-quality retrospective mammography dataset from Victoria, Australia to conduct detailed simulations of five potential AI-integrated screening pathways, and examine human-AI interaction effects to explore automation bias. Operating an AI reader as a second reader or as a high confidence filter improves current screening outcomes by 1.9–2.5% in sensitivity and up to 0.6% in specificity, achieving 4.6–10.9% reduction in assessments and 48–80.7% reduction in human reads.

。将AI读取器用作第二读取器或高可信度过滤器可将当前筛查结果的敏感性提高1.9–2.5%，特异性提高0.6%，评估减少4.6–10.9%，人类阅读减少48–80.7%。

Automation bias degrades performance in multi-reader settings but improves it for single-readers. This study provides insight into feasible approaches for AI-integrated screening pathways and prospective studies necessary prior to clinical adoption..

自动化偏差会降低多读取器设置中的性能，但会改善单读取器的性能。这项研究深入了解了AI综合筛查途径的可行方法以及临床采用前所需的前瞻性研究。。

IntroductionBreast cancer is the world’s most common cancer and a leading cause of cancer death in women1. BreastScreen Australia offers free mammographic screening targeted to women aged 50–74 years, with those over 40 years of age eligible to attend. Approximately 1 million women are screened annually, and the programme has achieved a reduction in mortality of 41–52% for screening participants and a 21% reduction in population-level breast cancer mortality2,3.

简介乳腺癌是世界上最常见的癌症，也是女性癌症死亡的主要原因1。澳大利亚BreastScreen为50-74岁的女性提供免费的乳房X光检查，40岁以上的女性有资格参加。每年大约有100万女性接受筛查，该计划使筛查参与者的死亡率降低了41-52%，人口乳腺癌死亡率降低了21%2,3。

However, there are challenges in accuracy, service experience, and efficiency.In 2020, ~59 per 10,000 participants were diagnosed with breast cancer and 16 per 10,000 participants were diagnosed with ductal carcinoma in situ (DCIS)2. Despite a process of independent double reading of all mammograms by radiologists, and a third arbitration read when there is discordance (henceforth called two readers with arbitration system), in 2020 ~368 per 10,000 participants were recalled for assessment and later determined not to have breast cancer (false positive).

但是，在准确性、服务体验和效率方面存在挑战。2020年，每10000名参与者中约有59名被诊断出患有乳腺癌，每10000名参与者中有16名被诊断出患有导管原位癌（DCIS）2。尽管放射科医生对所有乳房X线照片进行了独立的双重阅读，并且在不一致的情况下进行了第三次仲裁（以下称为仲裁系统的两名读者），但在2020年，每10000名参与者中有368人被召回进行评估，后来确定没有乳腺癌（假阳性）。

Also, ~18.6 per 10,000 participants aged 50–74 years (2015–2017) subsequently discovered they had an interval breast cancer before their next scheduled screen after receiving an all-clear result (false negative)2.Using artificial intelligence (AI) to help read mammograms has the potential to transform breast cancer screening by addressing the three key challenges of accuracy, service experience, and efficiency2.

此外，每10000名年龄在50-74岁（2015-2017年）的参与者中，约有18.6人在收到完全清楚的结果（假阴性）后，在下一次预定筛查之前发现他们患有间歇性乳腺癌。2.使用人工智能（AI）帮助阅读乳房X线照片有可能通过解决准确性，服务体验和效率这三个关键挑战来改变乳腺癌筛查2。

The evidence base for AI readers in breast cancer screening has been growing rapidly in recent years, with studies demonstrating the potential of AI to detect breast cancer on mammographic images with similar accuracy to radiologists4,5,6,7,8,9,10,11,12 and addressing key limitations of earlier concern13.

近年来，AI读者在乳腺癌筛查中的证据基础迅速增长，研究表明AI在乳腺X线照片上检测乳腺癌的潜力与放射科医生相似[4,5,6,7,8,9,10,11,12]，并解决了早期关注的关键局限性13。

Many of these studies evaluated the integration of.

许多这些研究评估了整合。

Data availability

数据可用性

Source data files are provided with this paper for all figures and tables. The non-transformed image and non-image data that established the ADMANI datasets were accessed under license agreement with BreastScreen Victoria. Further details about the ADMANI datasets are available in the data descriptor paper28.

本文提供了所有图形和表格的源数据文件。建立ADMANI数据集的未转换图像和非图像数据是根据与BreastScreen Victoria的许可协议访问的。有关ADMANI数据集的更多详细信息，请参阅数据描述符文件28。

The three datasets used as external validation are publicly available or available via request. The Chinese Mammography Dataset (CMMD) is publicly available from the following website: https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70230508. The Cohort of Screen-age Women - Case-control (CSAW-CC) dataset is available via request from the following website: https://snd.gu.se/en/catalogue/study/2021-204.

用作外部验证的三个数据集可公开获得或通过请求获得。中国乳腺X线摄影数据集（CMMD）可从以下网站公开获得：https://wiki.cancerimagingarchive.net/pages/viewpage.action?pageId=70230508.筛查年龄女性队列-病例对照（CSAW-CC）数据集可通过以下网站请求获得：https://snd.gu.se/en/catalogue/study/2021-204.

The BreastScreen Reader Assessment Strategy Australia (BREAST Australia) is available via request from the following website: https://breast-australia.sydney.edu.au/research/. Source data are provided with this paper..

澳大利亚BreastScreen读者评估策略（BREAST Australia）可通过以下网站索取：https://breast-australia.sydney.edu.au/research/.。。

Code availability

代码可用性

The code used for training the BRAIx AI reader is based on open-source algorithms and training techniques but is not able to be shared. We are required to protect potentially commercially valuable project intellectual property, which the source code constitutes, as part of our multi-institution agreement and grant obligations.

用于训练BRAIx AI阅读器的代码基于开源算法和训练技术，但无法共享。作为我们多机构协议和授予义务的一部分，我们需要保护源代码构成的潜在商业价值的项目知识产权。

The description of the model training procedure and models used are provided in the Methods section and can be implemented with open-source frameworks. The main conclusions drawn in our work relate to our AI reader simulation experiments. The code used to simulate the AI reader operating within the screening programme and validate the external results is available publicly43..

方法部分提供了模型训练过程和所用模型的描述，可以使用开源框架实现。我们工作中得出的主要结论与我们的AI阅读器模拟实验有关。用于模拟AI阅读器在筛选程序中运行并验证外部结果的代码可公开获得43。。

ReferencesWorld Cancer Research Fund. Breast cancer. https://www.wcrf.org/dietandcancer/breast-cancer/ (2021).Australian Institute of Health and Welfare. BreastScreen Australia Monitoring Report 2022 (Australian Institute of Health and Welfare, 2022).Morrell, S., Taylor, R., Roder, D. & Dobson, A. Mammography screening and breast cancer mortality in australia: an aggregate cohort study.

参考世界癌症研究基金会。乳腺癌。https://www.wcrf.org/dietandcancer/breast-cancer/（2021年）。澳大利亚健康与福利研究所。《2022年澳大利亚BreastScreen监测报告》（澳大利亚健康与福利研究所，2022年）。。

J. Med. Screen. 19, 26–34 (2012).Article .

J、 医学屏幕。19,26-34（2012）。。

Google Scholar

谷歌学者

Rodríguez-Ruiz, A. et al. Stand-alone artificial intelligence for breast cancer detection in mammography: comparison with 101 radiologistsdembrower. J. Natl. Cancer Inst. 111, 916–922 (2019).Article

Rodríguez-Ruiz，A.等人。乳腺X线摄影中乳腺癌检测的独立人工智能：与101位放射科医师Dembrower的比较。J、 纳特尔。癌症研究所111916-922（2019）。文章

Google Scholar

谷歌学者

Rodríguez-Ruiz, A. et al. Detection of breast cancer with mammography: effect of an artificial intelligence support system. Radiology 290, 305–314 (2019).Article

Rodríguez-Ruiz，A.等人。乳腺X线摄影检测乳腺癌：人工智能支持系统的作用。放射学290305-314（2019）。文章

Google Scholar

谷歌学者

McKinney, ScottMayer et al. International evaluation of an ai system for breast cancer screening. Nature 577, 89–94 (2020).Article

McKinney，ScottMayer等人。用于乳腺癌筛查的ai系统的国际评估。《自然》577,89-94（2020）。文章

ADS

广告

CAS

中科院

Google Scholar

谷歌学者

Kim, Hyo-Eun et al. Changes in cancer detection and false-positive recall in mammography using artificial intelligence: a retrospective, multireader study. Lancet Digit. Health 2, e138–e148 (2020).Article

Kim，Hyo Eun等人。使用人工智能的乳腺X线照相术中癌症检测和假阳性回忆的变化：一项回顾性多读者研究。柳叶刀数字。。文章

Google Scholar

谷歌学者

Ribli, Dezső., Horváth, A., Unger, Z., Pollner, P. éter & Csabai, István Detecting and classifying lesions in mammograms with deep learning. Sci. Rep. 8, 1–7 (2018).Article

里布利，迪兹。，Horváth，A.，Unger，Z.，Pollner，P。éter＆Csabai，István通过深度学习检测和分类乳房X线照片中的病变。科学。代表8，1-7（2018）。文章

CAS

中科院

Google Scholar

谷歌学者

Salim, M. et al. External evaluation of 3 commercial artificial intelligence algorithms for independent assessment of screening mammograms. JAMA Oncol. 6, 1581–1588 (2020).Article

Salim，M.等人。用于独立评估筛查乳房X线照片的3种商业人工智能算法的外部评估。JAMA Oncol。61581-1588（2020）。文章

Google Scholar

谷歌学者

Schaffter, T. et al. Evaluation of combined artificial intelligence and radiologist assessment to interpret screening mammograms. JAMA Netw. Open 3, e200265–e200265 (2020).Article

Schaffter，T.等人。人工智能和放射科医师联合评估对乳腺X线筛查的评估。JAMA网络。开放3，e200265–e200265（2020）。文章

Google Scholar

谷歌学者

Wu, N. et al. Deep neural networks improve radiologists’ performance in breast cancer screening. IEEE Trans. Med. Imag. 39, 1184–1194 (2020).Article

Wu，N.等人。深度神经网络改善了放射科医生在乳腺癌筛查中的表现。IEEE Trans。医学图像。391184-1194（2020）。文章

Google Scholar

谷歌学者

Frazer, HelenM. L., Qin, A. K., Pan, H. & Brotchie, P. Evaluation of deep learning-based artificial intelligence techniques for breast cancer detection on mammograms: results from a retrospective study using a breastscreen victoria dataset. J. Med. Imag. Rad. Oncol. 65, 529–537 (2021).Article .

弗雷泽，海伦。五十、 ，Qin，A.K.，Pan，H。＆Brotchie，P。基于深度学习的人工智能技术在乳腺X线照片上检测乳腺癌的评估：使用breastscreen victoria数据集进行回顾性研究的结果。J、 医学图像。放射性Oncol。65529-537（2021）。。

Google Scholar

谷歌学者

Freeman, K. et al. Use of artificial intelligence for image analysis in breast cancer screening programmes: systematic review of test accuracy. BMJ 374, n1872 (2021).Larsen, M. et al. Artificial intelligence evaluation of 122,969 mammography examinations from a population-based screening program.

。BMJ 374，n1872（2021）。Larsen，M.等人。基于人群的筛查计划对122969次乳房X光检查的人工智能评估。

Radiology 303, 502–511 (2022).Article .

。。

Google Scholar

谷歌学者

Yala, A., Schuster, T., Miles, R., Barzilay, R. & Lehman, C. A deep learning model to triage screening mammograms: a simulation study. Radiology 293, 38–46 (2019).Article

Yala，A.，Schuster，T.，Miles，R.，Barzilay，R。＆Lehman，C。分类筛查乳房X线照片的深度学习模型：模拟研究。放射学293,38-46（2019）。文章

Google Scholar

谷歌学者

Shoshan, Y. et al. Artificial intelligence for reducing workload in breast cancer screening with digital breast tomosynthesis. Radiology 303, 69–77 (2022).Article

Shoshan，Y.等人。使用数字乳腺断层合成减少乳腺癌筛查工作量的人工智能。放射学303,69-77（2022）。文章

Google Scholar

谷歌学者

Leibig, C. et al. Combining the strengths of radiologists and ai for breast cancer screening: a retrospective analysis. Lancet Digit. Health 4, e507–e519 (2022).Article

Leibig，C.等人。结合放射科医生和人工智能的优势进行乳腺癌筛查：回顾性分析。柳叶刀数字。健康4，e507–e519（2022）。文章

CAS

中科院

Google Scholar

谷歌学者

Lauritzen, A. D. et al. An artificial intelligence–based mammography screening protocol for breast cancer: Outcome and radiologist workload. Radiology 304, 41–49 (2022).Dembrower, K. et al. Effect of artificial intelligence-based triaging of breast cancer screening mammograms on cancer detection and radiologist workload: a retrospective simulation study.

Lauritzen，A.D.等人。基于人工智能的乳腺X线摄影筛查方案：结果和放射科医生的工作量。放射学304,41-49（2022）。Dembrower，K.等人。基于人工智能的乳腺癌筛查乳房X线照片分类对癌症检测和放射科医生工作量的影响：一项回顾性模拟研究。

Lancet Digit. Health 2, e468–e474 (2020).Article .

柳叶刀数字。健康2，e468-e474（2020）。。

Google Scholar

谷歌学者

Sharma, N. et al. Multi-vendor evaluation of artificial intelligence as an independent reader for double reading in breast cancer screening on 275,900 mammograms. BMC Cancer 23, 460 (2023).Article

Sharma，N.等人，《人工智能作为独立读取器在275900张乳腺X线照片上进行乳腺癌筛查的多供应商评估》。BMC癌症23460（2023）。文章

Google Scholar

谷歌学者

Lång, K. et al. Artificial intelligence-supported screen reading versus standard double reading in the mammography screening with artificial intelligence trial (masai): a clinical safety analysis of a randomised, controlled, non-inferiority, single-blinded, screening accuracy study. Lancet Oncol.

Lång，K.等人。人工智能支持屏幕阅读与人工智能乳房X线照相筛查试验（masai）中的标准双重阅读：一项随机，对照，非劣效，单盲，筛查准确性研究的临床安全性分析。柳叶刀Oncol。

24, 936–944 (2023).Article .

24936-944（2023）。。

Google Scholar

谷歌学者

Marinovich, M. L. et al. Artificial intelligence (AI) for breast cancer screening: BreastScreen population-based cohort study of cancer detection. EBioMedicine 90, 104498 (2023).Article

Marinovich，M.L.等人。用于乳腺癌筛查的人工智能（AI）：基于乳腺筛查人群的癌症检测队列研究。EBioMedicine 90104498（2023）。文章

CAS

中科院

Google Scholar

谷歌学者

Carter, S. M. et al. The ethical, legal and social implications of using artificial intelligence systems in breast cancer care. Breast 49, 25–32 (2020).Article

Carter，S.M.等人，《在乳腺癌护理中使用人工智能系统的伦理、法律和社会影响》。乳房49,25-32（2020）。文章

Google Scholar

谷歌学者

Byng, D. et al. Abstract ot3-18-03: the praim study: a prospective multicenter observational study of an integrated artificial intelligence system with live monitoring. Cancer Res. 83, OT3–18 (2023).Article

Byng，D.等人，《摘要ot3-18-03：praim研究：一项针对具有实时监测的综合人工智能系统的前瞻性多中心观察性研究》。癌症研究83，ot3-18（2023）。文章

Google Scholar

谷歌学者

Strand, F. Artificial intelligence in large-scale breast cancer screening (screentrustcad). ClinicalTrials.gov identifier: NCT04778670. Updated: 2023-03-14. Accessed: 2024-04-08. https://clinicaltrials.gov/study/NCT04778670.Al-Bazzaz, H., Janicijevic, M. & Strand, F. Reader bias in breast cancer screening related to cancer prevalence and artificial intelligence decision support—a reader study.

Strand，F。大规模乳腺癌筛查中的人工智能（screentrustcad）。ClinicalTrials.gov标识符：NCT04778670。更新日期：2023-03-14。访问时间：2024-04-08。https://clinicaltrials.gov/study/NCT04778670.Al-Bazzaz，H.，Janicijevic，M。＆Strand，F。与癌症患病率和人工智能决策支持相关的乳腺癌筛查中的读者偏见-读者研究。

Eur. Radiol. 34, 5415–5424 (2024).Goddard, K., Roudsari, A. & Wyatt, J. C. Automation bias: a systematic review of frequency, effect mediators, and mitigators. J. Am. Med. Inform. Assoc. 19, 121–127 (2012).Article .

欧洲放射性。345415-5424（2024）。Goddard，K.，Roudsari，A。＆Wyatt，J.C。自动化偏见：频率，效应介质和缓解因素的系统评价。J、 上午医疗通知。协会第19121-127号（2012年）。。

Google Scholar

谷歌学者

Frazer, HelenM. L. et al. Admani: annotated digital mammograms and associated non-image datasets. Radiol. Artif. Intell. 5, e220072 (2022).Article

弗雷泽，海伦。五十、 Admani等人：带注释的数字乳房X线照片和相关的非图像数据集。放射性。人工制品。因特尔。5，e220072（2022）。文章

Google Scholar

谷歌学者

Wilder, B., Horvitz, E. & Kamar, E. Learning to complement humans. In IJCAI (ed. Bessiere, C.) 1526–1533 (ijcai.org, 2020).He, K., Zhang, X., Ren, S. & Sun, J. Deep residual learning for image recognition. In Proc. IEEE Conference on Computer Vision and Pattern Recognition 770–778 (IEEE, 2016).Huang, G., Liu, Z., van der Maaten, L.

Wilder，B.，Horvitz，E。和Kamar，E。学习补充人类。在IJCAI（编辑贝西埃尔，C。）1526-1533（IJCAI.org，2020）。He，K.，Zhang，X.，Ren，S。＆Sun，J。用于图像识别的深度残差学习。在过程中。IEEE计算机视觉和模式识别会议770-778（IEEE，2016）。黄，G.，刘，Z.，范德马滕，L。

& Weinberger, K. Q. Densely connected convolutional networks. In Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 7 (IEEE, 2017).Wang, Q. et al. ECA-Net: efficient channel attention for deep convolutional neural networks. In CVPR 11531–11539 (Computer Vision Foundation / IEEE, 2020).Tan, M.

&Weinberger，K.Q。密集连接的卷积网络。在过程中。IEEE计算机视觉和模式识别会议（CVPR）7（IEEE，2017）。Wang，Q。等。ECA网络：深度卷积神经网络的有效通道注意。在CVPR 11531–11539（计算机视觉基金会/IEEE，2020）中。谭，M。

& Le, Q. Efficientnet: rethinking model scaling for convolutional neural networks. In Proc. 36th International Conference on Machine Learning 6105–6114 (PMLR, 2019).Szegedy, C., Ioffe, S., Vanhoucke, V. & Alemi, A. A. Inception-v4, inception-resnet and the impact of residual connections on learning.

&Le，Q。Efficientnet：重新思考卷积神经网络的模型缩放。在过程中。第36届国际机器学习会议6105-6114（PMLR，2019）。Szegedy，C.，Ioffe，S.，Vanhoucke，V。＆Alemi，A.A。Inception-v4，Inception resnet和剩余连接对学习的影响。

In Thirty-first AAAI Conference on Artificial Intelligence (AAAI Press, 2017).Chollet, F. Xception: deep learning with depthwise separable convolutions. In Proc. IEEE Conference on Computer Vision and Pattern Recognition (CVPR) 7 (IEEE, 2017).Liu, Z. et al. A convnet for the 2020s. In Proc. IEEE/CVF Conference on Computer Vision and Pattern Recognition 11976–11986 (IEEE, 2022).Chen, Y.

在第三十一届AAAI人工智能会议上（AAAI Press，2017）。。在过程中。IEEE计算机视觉和模式识别会议（CVPR）7（IEEE，2017）。Liu，Z.等人，《20世纪20年代的变革》。在过程中。IEEE/CVF计算机视觉和模式识别会议11976–11986（IEEE，2022）。陈，Y。

et al. Multi-view local co-occurrence and global consistency learning improve mammogram classification generalisation. In Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 3–13 (Cham, Springer, 2022).Wang, C. et al. Knowledge distillation to ensemble global and interpretable prototype-based mammogram classification models.

多视图局部共现和全局一致性学习改善了乳房X线照片分类的普遍性。医学图像计算和计算机辅助干预–MICCAI 2022 3–13（Cham，Springer，2022）。Wang，C.等人。知识提炼，以集成基于全局和可解释原型的乳腺X线分类模型。

In Medical Image Computing and Computer Assisted Intervention – MICCAI 2022 14–24 (Springer, 2022.

医学图像计算和计算机辅助干预-MICCAI 2022 14-24（Springer，2022）。

Google Scholar

谷歌学者

Kwok, C. F. & Elliott, M. S. Braix-project/retrospective-cohort-study: v3.0.0. Zenodo https://doi.org/10.5281/zenodo.12633016 (2024).Youden, W. J. Index for rating diagnostic tests. Cancer 3, 32–35 (1950).Article

Kwok，C.F。和Elliott，M.S.Braix项目/回顾性队列研究：v3.0.0。泽诺多https://doi.org/10.5281/zenodo.12633016（2024年）。Youden，W.J。评估诊断测试的索引。癌症3,32-35（1950）。文章

CAS

中科院

Google Scholar

谷歌学者

Cancer Australia. Guidance for the management of early breast cancer: recommendations and practice points. Cancer Australia (2020).Lee, J. M. et al. Breast cancer risk, worry, and anxiety: effect on patient perceptions of false-positive screening results. Breast 50, 104–112 (2020).Article .

澳大利亚癌症。早期乳腺癌管理指南：建议和实践要点。澳大利亚癌症（2020）。Lee，J.M.等人，《乳腺癌风险、担忧和焦虑：对患者对假阳性筛查结果认知的影响》。。。

Google Scholar

谷歌学者

McNemar, Q. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika 12, 153–157 (1947).Article

McNemar，Q。注意相关比例或百分比之间差异的抽样误差。心理测量学12153-157（1947）。文章

CAS

中科院

Google Scholar

谷歌学者

Trajman, A. & Luiz, R. R. Mcnemar χ2 test revisited: comparing sensitivity and specificity of diagnostic examinations. Scand. J. Clin. Lab. Invest. 68, 77–80 (2008).Article

Trajman，A。＆Luiz，R。R。Mcnemarχ2检验重新审视：比较诊断检查的敏感性和特异性。斯堪的纳维亚。J、 临床。实验室投资。68，77-80（2008）。文章

CAS

中科院

Google Scholar

谷歌学者

Kim, S. & Lee, W. Does mcnemar’s test compare the sensitivities and specificities of two diagnostic tests? Stat. Methods Med. Res. 26, 142–154 (2017).Article

Kim，S.＆Lee，W。mcnemar的测试是否比较了两种诊断测试的敏感性和特异性？《统计方法医学》第26142-154号决议（2017年）。文章

MathSciNet

MathSciNet

Google Scholar

谷歌学者

Download referencesAcknowledgementsThe authors would like to acknowledge Rita Butera, Luke Neill, and Georgina Marr from BreastScreen Victoria and the leadership and staff of St Vincent’s Hospital Melbourne for their support of the project. The authors would like to acknowledge Katrina Kunicki, Anne Johnston, Colleen Elso, Elizabeth Campbell and Tom Kay of St Vincents Institute of Medical Research (SVI) for their extensive help and support with many aspects of this project.

下载参考文献致谢作者要感谢BreastScreen Victoria的Rita Butera，Luke Neill和Georgina Marr以及墨尔本圣文森特医院的领导和工作人员对该项目的支持。作者要感谢圣文森特医学研究所（SVI）的卡特里娜·库尼基（Katrina Kunicki）、安妮·约翰斯顿（Anne Johnston）、科琳·埃尔索（Colleen Elso）、伊丽莎白·坎贝尔（Elizabeth Campbell）和汤姆·凯（Tom Kay），感谢他们在该项目的许多方面提供的广泛帮助和支持。

The authors thank Wayne Crismani for thoughtful comments on the manuscript. This work is supported by funding from the Australian Government under the Medical Research Future Fund Grant (MRFAI000090) for the Transforming Breast Cancer Screening with Artificial Intelligence (BRAIx) Project awarded to H.M.L.F., D.J.M., P.B., J.F.L., J.L.H., G.C.

作者感谢Wayne Crismani对稿件的深思熟虑的评论。这项工作得到了澳大利亚政府在医学研究未来基金赠款（MRFAI000090）下的资助，该赠款用于通过人工智能转化乳腺癌筛查（BRAIx）项目，该项目授予H.M.L.F.，D.J.M.，P.B.，J.F.L.，J.L.H.，G.C。

and a National Health and Medical Research Council Investigator Grant (GNT1195595) awarded to D.J.M. This work is also supported by a Ramaciotti Health Investment Grant awarded to D.J.M. and funding from a Royal Australian and New Zealand College of Radiologists Clinical Research Grant and the St Vincent’s Hospital Melbourne Research Endowment Fund awarded to HMLF.

以及授予D.J.M.的国家卫生与医学研究委员会研究者补助金（GNT1195595）。这项工作还得到了授予D.J.M.的拉马西奥蒂健康投资补助金以及澳大利亚和新西兰皇家放射科医师学院临床研究补助金和授予HMLF的圣文森特医院墨尔本研究捐赠基金的资助。

The funders had no role in the work or decision to publish. SVI provided significant in-kind support comprising IT support, infrastructure, and GPUs used for numerical calculations in this paper enabled via funding from the Victorian State Government Operational Infrastructure Support Programme to St Vincent’s Institute of Medical Research.

资助者在出版的工作或决定中没有任何作用。SVI提供了大量实物支持，包括IT支持、基础设施和本文中用于数值计算的GPU，这些支持是通过维多利亚州政府运营基础设施支持计划向圣文森特医学研究所提供的资金实现的。

HMLF acknowledges the generous support of the Medical Device Partnering Programme for enabling the prior foundational research to be undertaken. D.J.M. acknowledges generous support from Paul Holyoake and Marg Downey.Author informationAuthor notesT.

HMLF感谢医疗器械合作计划的慷慨支持，以便能够进行先前的基础研究。D、 J.M.感谢Paul Holyoake和Marg Downey的慷慨支持。。

PubMed Google ScholarCarlos A. Peña-SolorzanoView author publicationsYou can also search for this author in

PubMed Google ScholarCarlos A.PeñA-SolorzanoView作者出版物您也可以在

PubMed Google ScholarChun Fung KwokView author publicationsYou can also search for this author in

PubMed Google ScholarChun Fung KwokView作者出版物您也可以在

PubMed Google ScholarMichael S. ElliottView author publicationsYou can also search for this author in

PubMed Google Scholarmamichael S.ElliottView作者出版物您也可以在

PubMed Google ScholarYuanhong ChenView author publicationsYou can also search for this author in

PubMed Google ScholaryHuanhong ChenView作者出版物您也可以在

PubMed Google ScholarChong WangView author publicationsYou can also search for this author in

PubMed Google ScholarChong WangView作者出版物您也可以在

PubMed Google ScholarJocelyn F. LippeyView author publicationsYou can also search for this author in

PubMed Google ScholarJocelyn F.LippeyView作者出版物您也可以在

PubMed Google ScholarJohn L. HopperView author publicationsYou can also search for this author in

PubMed谷歌学者John L.HopperView作者出版物您也可以在

PubMed Google ScholarPeter BrotchieView author publicationsYou can also search for this author in

PubMed Google ScholarPeter BrotchieView作者出版物您也可以在

PubMed Google ScholarGustavo CarneiroView author publicationsYou can also search for this author in

PubMed Google ScholarGustavo CarneiroView作者出版物您也可以在

PubMed Google ScholarDavis J. McCarthyView author publicationsYou can also search for this author in

PubMed Google ScholarDavis J.McCarthyView作者出版物您也可以在

PubMed Google ScholarConsortiaThe BRAIx TeamOsamah Al-Qershi, Peter Brotchie, Gustavo Carneiro, Yuanhong Chen, Michael S. Elliott, Samantha K. Fox, Helen M. L. Frazer, Brendan Hill, John L. Hopper, Ravishankar Karthik, Katrina Kunicki, Chun Fung Kwok, Shuai Li, Jocelyn F. Lippey, Enes Makalic, Davis J.

PubMed谷歌学术研讨会BRAIx TeamOsamah Al Qershi，Peter Brotchie，Gustavo Carneiro，Yuanhong Chen，Michael S.Elliott，Samantha K.Fox，Helen M.L.Frazer，Brendan Hill，John L.Hopper，Ravishankar Karthik，Katrina Kunicki，Chun Fung Kwok，Shuai Li，Jocelyn F.Lippey，Enes Makalic，Davis J。

McCarthy, Tuong L. Nguyen, Carlos A. Peña-Solorzano, Prabhathi Basnayake Ralalage, Daniel Schmidt, Chong Wang & Prue C. WeidemanContributionsConceptualisation: H.M.L.F., D.J.M., J.F.L., J.L.H. and G.C.; Methodology: H.M.L.F., C.A.P.-S., C.F.K., M.S.E., Y.C., C.W., P.B., G.C., D.J.M.; AI model and code development: C.A.P.-S., C.F.K., M.S.E., Y.C.

麦卡锡，Tuong L.Nguyen，Carlos A.PeñA-Solorzano，Prabhathi Basnayake Ralalage，Daniel Schmidt，Chong Wang和Prue C.WeidemanContributions概念化：H.M.L.F.，D.J.M.，J.F.L.，J.L.H.和G.C。；方法学：H.M.L.F.，C.A.P.-S.，C.F.K.，M.S.E.，Y.C.，C.W.，P.B.，G.C.，D.J.M。；AI模型和代码开发：C.A.P.-S.，C.F.K.，M.S.E.，Y.C。

and C.W.; Dataset development: H.M.L.F., M.S.E., C.A.P.-S. and C.F.K.; Data analysis: C.A.P.-S., C.F.K. and M.S.E.; Resources: D.J.M., H.M.L.F., P.B., J.F.L., J.L.H. and G.C.; Data preparation: M.S.E., B.H. and R.K.; Data interpretation: all authors; Results interpretation: all authors; Writing—original draft, review and editing: C.A.P.-S., C.F.K., M.S.E., H.M.L.F., D.J.M.

和C.W。；数据集开发：H.M.L.F.，M.S.E.，C.A.P.-S.和C.F.K。；数据分析：C.A.P.-S.，C.F.K.和M.S.E。；资源：D.J.M.，H.M.L.F.，P.B.，J.F.L.，J.L.H.和G.C。；数据准备：M.S.E.，B.H.和R.K。；数据解释：所有作者；结果解释：所有作者；撰写原稿，审查和编辑：C.A.P.-S.，C.F.K.，M.S.E.，H.M.L.F.，D.J.M。

and G.C.; Supervision: H.M.L.F., D.J.M. and G.C.; Project administration: H.M.L.F., D.J.M. and G.C.; Funding acquisition: H.M.L.F., D.J.M., P.B., J.F.L., J.L.H. and G.C.Corresponding authorCorrespondence to.

和G.C。；监督：H.M.L.F.，D.J.M.和G.C。；项目管理：H.M.L.F.，D.J.M.和G.C。；资金获取：H.M.L.F.，D.J.M.，P.B.，J.F.L.，J.L.H.和G.C.对应作者回复。

Helen M. L. Frazer.Ethics declarations

。道德宣言

Competing interests

相互竞争的利益

P.B. is an employee of annalise.ai. C.W., Y.C., D.J.M., M.S.E., H.M.L.F. and G.C. are inventors on a patent, 'WO2024044815—Improved classification methods for machine learning', a model used in versions of the BRAIx AI reader. The remaining authors declare no competing interests.

P、 B.是annalise.ai.C.W.，Y.C.，D.J.M.，M.S.E.，H.M.L.F.和G.C.的员工，是“WO2024044815改进的机器学习分类方法”专利的发明人，该专利用于BRAIx ai阅读器的版本。其余作者声明没有利益冲突。

Peer review

同行评审

Peer review information

同行评审信息

Nature Communications thanks the anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Nature Communications感谢匿名审稿人对这项工作的同行评审做出的贡献。同行评审文件可用。

Additional informationPublisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Supplementary informationSupplementary InformationPeer Review FileReporting SummarySource dataSource DataRights and permissions

Additional informationPublisher的注释Springer Nature在已发布的地图和机构隶属关系中的管辖权主张方面保持中立。补充信息补充信息同行评审文件报告摘要源数据源数据权限

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material.

开放获取本文是根据知识共享署名非商业性NoDerivatives 4.0国际许可证授权的，该许可证允许以任何媒介或格式进行任何非商业性使用，共享，分发和复制，只要您对原始作者和来源给予适当的信任，提供知识共享许可证的链接，并指出您是否修改了许可材料。

You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.

根据本许可证，您无权共享源自本文或其部分的改编材料。本文中的图像或其他第三方材料包含在文章的知识共享许可证中，除非该材料的信用额度中另有说明。如果材料未包含在文章的知识共享许可中，并且您的预期用途不受法律法规的许可或超出许可用途，则您需要直接获得版权所有者的许可。

To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/..

要查看此许可证的副本，请访问http://creativecommons.org/licenses/by-nc-nd/4.0/..

Reprints and permissionsAbout this articleCite this articleFrazer, H.M.L., Peña-Solorzano, C.A., Kwok, C.F. et al. Comparison of AI-integrated pathways with human-AI interaction in population mammographic screening for breast cancer.

转载和许可本文引用了这篇文章Frazer，H.M.L.，Peña-Solorzano，C.a.，Kwok，C.F。等人在乳腺癌人群乳腺X线筛查中AI整合途径与人类AI相互作用的比较。

Nat Commun 15, 7525 (2024). https://doi.org/10.1038/s41467-024-51725-8Download citationReceived: 25 September 2023Accepted: 14 August 2024Published: 30 August 2024DOI: https://doi.org/10.1038/s41467-024-51725-8Share this articleAnyone you share the following link with will be able to read this content:Get shareable linkSorry, a shareable link is not currently available for this article.Copy to clipboard.

《国家公社》157525（2024）。https://doi.org/10.1038/s41467-024-51725-8Download引文接收日期：2023年9月25日接收日期：2024年8月14日发布日期：2024年8月30日OI：https://doi.org/10.1038/s41467-024-51725-8Share本文与您共享以下链接的任何人都可以阅读此内容：获取可共享链接对不起，本文目前没有可共享的链接。复制到剪贴板。

Provided by the Springer Nature SharedIt content-sharing initiative

由Springer Nature SharedIt内容共享计划提供

CommentsBy submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.

评论通过提交评论，您同意遵守我们的条款和社区指南。。