[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-352322-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-352322-en":130},{"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","improving-generalization-of-polyp-detection-via-conditional-stylegan-augmented-training","Improving generalization of polyp detection via conditional StyleGAN augmented training","","Colorectal cancer outcomes depend on early diagnosis, yet colonoscopy screening misses subtle flat or serrated lesions, compounded by limited high-quality annotated datasets for computer-aided detection. The study uses a conditional StyleGAN to synthesize high-resolution colorectal neoplasm images, built from over 150,000 images from diverse public sources. YOLOv5 models trained with this synthetic data achieve high fidelity, improve mean Average Precision from 0.86 to 0.93, reduce generalization gaps on external sets, and raise recall for challenging flat and depressed lesions from 0.72 to 0.87.",{"@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/improving-generalization-of-polyp-detection-via-conditional-stylegan-augmented-training/352322/",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/improving-generalization-of-polyp-detection-via-conditional-stylegan-augmented-training/352322.png","ImageObject",300,407,{"name":42,"@type":43},"Ivy","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-25","2026-09-22",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",9,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"Why is improving polyp detection generalization important for colorectal cancer screening?","Question",{"text":63,"@type":64},"Colonoscopy screening has significant miss rates, especially for subtle flat or serrated lesions. Improving generalization helps AI-assisted detection remain reliable across diverse clinical scenarios.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does conditional StyleGAN augmented training address dataset limitations?",{"text":68,"@type":64},"The method synthesizes high-resolution images of colorectal neoplasms using a conditional StyleGAN trained with over 150,000 images aggregated from public datasets, reducing scarcity and imbalance in annotated data.",{"name":70,"@type":61,"acceptedAnswer":71},"What improvements were observed when using the synthetic data to train YOLOv5?",{"text":72,"@type":64},"Hybrid augmentation increased mean Average Precision from 0.86 to 0.93 on internal testing and reduced the generalization gap on independent external validation sets, with recall for challenging flat and depressed lesions rising from 0.72 to 0.87.","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},352322,1790600319,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,119,122,126],{"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":55,"doc_module":4,"doc_module_name":25,"category_name":116,"show_sort_weight":117,"slug":118},"Religion & Spirituality",20,"religion-spirituality",{"id":117,"doc_module":4,"doc_module_name":25,"category_name":120,"show_sort_weight":117,"slug":121},"World Cup","world-cup",{"id":123,"doc_module":4,"doc_module_name":25,"category_name":124,"show_sort_weight":123,"slug":125},10,"Lifestyle","lifestyle",{"id":127,"doc_module":4,"doc_module_name":25,"category_name":128,"show_sort_weight":97,"slug":129},19,"General","general",{"code":4,"msg":82,"data":131},{"doc_id":79,"user_id":132,"nickname":42,"user_avatar":133,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":139,"language":140,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":12,"update_tm":144,"read_time":145,"preview_page_count":33,"preview_urls":146},549758252649,"https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819","npj | digital medicine Article  \n\n| Published in partnership with Seoul National University Bundang Hospital |  | |\n| --- | --- | --- |\n| [https://doi.org/10.1038/s41746-025-02293-8](https://doi.org/10.1038/s41746-025-02293-8) |  |  |\n| Improving generalization of polyp detection via conditional StyleGAN augmented training\u003Cbr> Check for updates |  |  |\n| Yilin Lin1, Cong Huang2, Hairui Tian2, Bing Yang3,4, Tingting Deng2, Yu Pan5 , Hao Wang6  & Xu Li6  |  |  |\n| Colorectal cancer outcomes are critically dependent on early diagnosis, yet colonoscopy screening suffers from signiﬁcant miss rates, particularly for subtle ﬂat or serrated lesions. Although artiﬁcial intelligence holds promise for computer-aided detection (CADe), system performance is frequently bottlenecked by the scarcity and imbalance of high-quality annotated datasets. To address this, we employed a conditional StyleGAN architecture to synthesize high-resolution images of colorectal neoplasms, leveraging over 150,000 images aggregated from diverse public datasets. When utilized to train YOLOv5 detection models, this synthetic data demonstrated high ﬁdelity and signiﬁcantly enhanced diagnostic performance. Hybrid augmentation improved the mean Average Precision from 0.86 to 0.93 on internal testing and markedly reduced the generalization gap on independent external validation sets. Crucially, recall for challengingﬂat and depressed lesions rose from0.72to0.87. These ﬁndings indicate that generative augmentation effectively strengthens model robustness and generalization across diverse clinical scenarios. While currently limited to still imagery, this strategy provides a scalable solution to data limitations, potentially elevating the standard of AI-assisted endoscopic surveillance. |  |  |\n| Colorectal cancer (CRC) represents a formidable global public health challenge, ranking as the third most diagnosed cancer and the second leading cause of cancer mortality worldwide according to GLOBOCAN 2022 estimates, with nearly 2.0 million new cases and over 900,000 deaths annually. This burden is marked by signiﬁcant socioeconomic disparities; while high-income nations historically show the highest incidence rates linked to Westernized lifestyles, they are now experiencing a stabilization or decline due to effective screening, whereas rates are alarmingly escalating in developing regions where limited healthcare access leads to higher mortality. This epidemiological landscape translates into a staggering economic and societal strain, encompassing not only vast direct medical costs but also profound indirect costs from lost productivity and premature death, which are exponentially higher for late-stage versus early-stage disease. This stark correlation between diagnostic stage, patient survival, and economic burden creates a compelling and urgent imperative for global strategies focused on prevention and early detection Fig. 1. | The cornerstone of CRC prevention lies in its well-deﬁned and typically indolent natural history, which provides a crucial window for intervention. The majority of sporadic CRCs develop through the multi-step adenoma carcinoma sequence, shown in Fig. 2 a process where normal colonic epithelium transforms into a benign adenomatous polyp that, over a span of 10 to 15 years, accumulates a series of genetic and epigenetic alterations, ultimately leading to invasive cancer1,2. This progression is driven by distinct molecular pathways, primarily the Chromosomal Instability (CIN) pathway, characterized by mutations in genes like APC, KRAS, and TP53, and the Microsatellite Instability(MSI)pathway, which results from adeﬁcient DNA mismatch repair system3,4. This protracted, multi-stage evolution from a detectable, benign precursor to a malignant tumor is the fundamental biological rationale for screening programs, as the removal of adenomas via polypectomy effectively intercepts this sequence and prevents the development of cancer.\u003Cbr>The clinical imperati","cbCaioAzGpJF81kH","https://ap.wps.com/l/cbCaioAzGpJF81kH","pdf",3225038,14,"English","# Introduction\n## Clinical need for early colorectal cancer detection\n## Biological rationale for screening and stage-dependent outcomes\n# Generative augmentation for polyp detection","[{\"question\":\"Why is improving polyp detection generalization important for colorectal cancer screening?\",\"answer\":\"Colonoscopy screening has significant miss rates, especially for subtle flat or serrated lesions. Improving generalization helps AI-assisted detection remain reliable across diverse clinical scenarios.\"},{\"question\":\"How does conditional StyleGAN augmented training address dataset limitations?\",\"answer\":\"The method synthesizes high-resolution images of colorectal neoplasms using a conditional StyleGAN trained with over 150,000 images aggregated from public datasets, reducing scarcity and imbalance in annotated data.\"},{\"question\":\"What improvements were observed when using the synthetic data to train YOLOv5?\",\"answer\":\"Hybrid augmentation increased mean Average Precision from 0.86 to 0.93 on internal testing and reduced the generalization gap on independent external validation sets, with recall for challenging flat and depressed lesions rising from 0.72 to 0.87.\"}]","Improving generalization of polyp detection via conditional StyleGAN augmented training | PDF",1790098950,35,[147,151,155,159],{"page_no":22,"url":148,"width":149,"height":150},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/537f36ef5df27ba0e17fbab45e95732d.webp",983,1305,{"page_no":26,"url":152,"width":153,"height":154},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/c4c7e60ca99a9d0754807786c9487f71.webp",982,1304,{"page_no":30,"url":156,"width":157,"height":158},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/c6571bc31837613a8204e427f17798f4.webp",981,1303,{"page_no":33,"url":160,"width":161,"height":162},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/5c518857da7932169f20f42b636f44ee.webp",980,1301]