[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-350059-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-350059-en":131},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","diagnostic-accuracy-fairness-and-clinical-implementation-of-ai-for-breast-cancer-screening-results-of-multicenter-retrospective-and-prospective-technical-feasibility-studies","Diagnostic accuracy, fairness and clinical implementation of AI for breast cancer screening: results of multicenter retrospective and prospective technical feasibility studies","","Evaluates Google’s mammography AI system (version 1.2) for breast cancer screening across retrospective and prospective phases. The retrospective component used 115,973 mammograms from five NHS screening services with 39-month follow-up, comparing AI sensitivity and specificity against first readers under a 5% noninferiority margin. AI showed superior sensitivity (0.541 vs 0.437) and noninferior specificity (0.943 vs 0.952), increased cancer detection, and reduced recalls, with no systematic demographic disparities. Prospective deployment demonstrated technical feasibility but required threshold recalibration due to distribution shift, supporting adaptive calibration and continuous monitoring for safety and equity.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/diagnostic-accuracy-fairness-and-clinical-implementation-of-ai-for-breast-cancer-screening-results-of-multicenter-retrospective-and-prospective-technical-feasibility-studies/350059/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/diagnostic-accuracy-fairness-and-clinical-implementation-of-ai-for-breast-cancer-screening-results-of-multicenter-retrospective-and-prospective-technical-feasibility-studies/350059.png","ImageObject",300,407,{"name":42,"@type":43},"\tCallum ","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-11-14","2026-09-22",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",16,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What AI system and version were evaluated for breast cancer screening?","Question",{"text":63,"@type":64},"The study evaluated Google’s mammography AI system, version 1.2, across retrospective and prospective phases.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How did the AI perform compared with first readers in the retrospective study?",{"text":68,"@type":64},"AI achieved superior sensitivity (0.541 vs 0.437) and noninferior specificity (0.943 vs 0.952), using a 5% noninferiority margin.",{"name":70,"@type":61,"acceptedAnswer":71},"What did prospective deployment show and what operational adjustment was needed?",{"text":72,"@type":64},"Prospective deployment confirmed technical feasibility, but a distribution shift required threshold recalibration to maintain appropriate operation.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},350059,1790641981,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":140,"language":141,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":12,"update_tm":145,"read_time":146,"preview_page_count":33,"preview_urls":147},137451211410,"https://ap-avatar.wpscdn.com/avatar/2000bb0a9246f588df?x-image-process=image/resize,m_fixed,w_180,h_180&k=1786362646172706240","nature cancer  \nArticle [https://doi.org/10.1038/s43018-026-01127-0](https://doi.org/10.1038/s43018-026-01127-0)  \nDiagnostic accuracy, fairness and clinical implementation of AI for breast cancer screening: results of multicenter retrospective and prospective technical feasibility studies  \nReceived: 3 October 2024  \nAccepted: 16 January 2026  \n\n| Published online: 10 March 2026 |\n| --- |\n|  Check for updates |\n\nA list of authors and their affiliations appears atthe end of the paper  \nArtificial intelligence (AI) promises to enhance breast cancer screening.  \nHere we evaluated Google’s mammography AI system (version 1.2) across two phases: a retrospective study using 115,973 mammograms from five National Health Service screening services with 39-month follow-up and prospective noninterventional feasibility deployment at 12 sites (9,266 cases). The primary endpoint was AI sensitivity and specificity versus first  \nreader using a 5% noninferiority margin. The secondary endpoints were performance versus second or consensus readers and breast-level analyses. Retrospectively, AI achieved superior sensitivity (0.541 versus 0.437 for first reader, P \u003C 0.001) and noninferior specificity (0.943 versus 0.952, P \u003C 0.001). Cancer detection rate increased from 7.54 to 9.33 per 1,000 women, with AI detecting 25.0% of interval cancers. Performance was particularly strong for first screens (39.3% fewer recalls, 8.8% higher detection) and invasive cancers. No systematic demographic disparities were observed. Simulated second-reader replacement reduced reading time by 32% while increasing detection by 17.7%. Prospective deployment confirmed technical feasibility but revealed a distribution shift requiring threshold recalibration. Implementation requires adaptive calibration and continuous monitoring to ensure safety and equity.  \nBreast cancer screening offers a compelling clinical use case forartificial intelligence (AI) in healthcare. It is hoped that AI can improve the quality and consistency of screening while improving cost effectiveness and addressing global radiologist workforce shortages. A number of retrospective studies have demonstrated that AI can perform at least on a par with specialists in different settings1 and new evidence from interventional studies is growing2–4. Despite this, the use of modern AI in breast screening programs has not been widely adopted to date.  \nSeveral crucial pieces of evidence that thoroughly assess benefits and harms are still required to catalyze widespread adoption5. These include a better understanding of the spectrum of disease  \ndetected by AI, understanding the impact on interval cancers in large and diverse populations, ensuring safety and equity, careful exploration of human–computer interaction and implementation science to understand practical considerations when deploying AI at scale.  \nGlobal breast screening workflows can be divided into single-read (primarily United States) and double-read (other countries) workflows6. Double-read screening involves two readers separately reviewing each mammogram, where the second reader can be blinded orunblinded to the first reader’s decision. An arbitration panel is invoked incasesof reader disagreement and sometimes for all suspected cancer cases. In the double-read setting, there are at least three potential AI  \n[e-mail:](e-mail: christopher.kelly1@nhs.net)[ christopher.kelly1@nhs.net](e-mail: christopher.kelly1@nhs.net); [sshetty@google.com](sshetty@google.com); [h.ashrafian@imperial.ac.uk](h.ashrafian@imperial.ac.uk); [deborah.cunningham1@nhs.net](deborah.cunningham1@nhs.net)  \na Retrospective study of five di􀀋erent screening services  \nTotal study size:  \n125,000 women  \nLeicester  \nWest London  \nCornwall Oxford South West London  \nb Fairness and safety  \nDetailed subgroup analysis across clinical, social and demographic factors  \nOperating point selection performed for each screening service individually, with subsequent drift analysis  \n1.  \n2","cbCaij9MVXNwpYG2","https://ap.wps.com/l/cbCaij9MVXNwpYG2","pdf",12080041,26,"English","# Introduction\n## Rationale for AI in breast screening\n## Evidence gaps for adoption\n# Study design and workflows\n## Single-read vs double-read screening\n## Reader arbitration and AI integration options\n# Results: retrospective evaluation\n## Primary and secondary endpoints\n## Sensitivity, specificity, and detection metrics\n## Demographic disparity assessment\n# Results: prospective feasibility study\n## Deployment workflow and technical feasibility\n## Distribution shift and threshold recalibration\n# Implementation considerations\n## Adaptive calibration and continuous monitoring for safety and equity","[{\"question\":\"What AI system and version were evaluated for breast cancer screening?\",\"answer\":\"The study evaluated Google’s mammography AI system, version 1.2, across retrospective and prospective phases.\"},{\"question\":\"How did the AI perform compared with first readers in the retrospective study?\",\"answer\":\"AI achieved superior sensitivity (0.541 vs 0.437) and noninferior specificity (0.943 vs 0.952), using a 5% noninferiority margin.\"},{\"question\":\"What did prospective deployment show and what operational adjustment was needed?\",\"answer\":\"Prospective deployment confirmed technical feasibility, but a distribution shift required threshold recalibration to maintain appropriate operation.\"}]","Diagnostic accuracy, fairness and clinical implementation of AI for breast cancer screening: results of multicenter retrospective and prospective technical feasibility studies | PDF",1790086964,66,[148,152,156,160],{"page_no":22,"url":149,"width":150,"height":151},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/a1ddd4f21d28f8fc163c2057b379f79f.webp",983,1305,{"page_no":26,"url":153,"width":154,"height":155},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/42458748af7a6941af158448ceea2582.webp",982,1304,{"page_no":30,"url":157,"width":158,"height":159},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/424ce445cb074510e539ce0c08290bf2.webp",981,1303,{"page_no":33,"url":161,"width":162,"height":163},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/587bec2a2e1f2c3201500af93087438b.webp",980,1301]