[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-344456-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-344456-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","boosting-pathology-foundation-models-via-few-shot-prompt-tuning-for-rare-cancer-subtyping","Boosting pathology foundation models via few-shot prompt-tuning for rare cancer subtyping","","Rare cancers account for 20–25% of malignancies yet suffer major diagnostic difficulties because expert availability is limited and reference cases are scarce. While pathology vision-language models can perform zero-shot for common cancers, their performance on rare cancer subtyping remains inadequate. The proposed PathPT framework leverages vision-language foundation models with spatially aware visual aggregation and task-specific prompt tuning, converting WSI-level supervision into tile-level guidance for tumor localization and cross-modal reasoning. Experiments across rare and common datasets show consistent improvements in subtyping accuracy and cancerous region grounding under data-scarce conditions.",{"@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/boosting-pathology-foundation-models-via-few-shot-prompt-tuning-for-rare-cancer-subtyping/344456/",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/boosting-pathology-foundation-models-via-few-shot-prompt-tuning-for-rare-cancer-subtyping/344456.png","ImageObject",300,407,{"name":42,"@type":43},"Valentina","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-11-15","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},"Why are rare cancer subtyping tasks challenging in practice?","Question",{"text":63,"@type":64},"Rare cancers are collectively common but individually uncommon, and they lack clinical expertise and annotated reference cases. This limits reliable training signals and makes accurate, timely subtyping difficult.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"What limitations exist in adapting vision-language foundation models using the current MIL paradigm?",{"text":68,"@type":64},"Existing MIL approaches focus mainly on visual features and ignore cross-modal reasoning from the textual encoder. They also rely on slide-level supervision, providing weak spatial guidance for fine-grained region-specific patterns.",{"name":70,"@type":61,"acceptedAnswer":71},"How does PathPT improve rare cancer subtyping under few-shot or data-scarce conditions?",{"text":72,"@type":64},"PathPT introduces spatially-aware visual aggregation and task-specific prompt tuning to harness vision-language prior knowledge. It converts WSI-level supervision into fine-grained tile-level guidance, improving both subtyping accuracy and tumor region grounding.","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},344456,1790211538,{"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},13056703020460,"https://ap-avatar.wpscdn.com/avatar/be000253dac470eee5d?_k=1778207105932848923","Article [https://doi.org/10.1038/s41467-026-71715-2](https://doi.org/10.1038/s41467-026-71715-2)  \nBoosting pathology foundation models via few-shot prompt-tuning for rare cancer subtyping  \nReceived: 30 August 2025  \n\n| Accepted: 26 March 2026 |\n| --- |\n| |\n| Check for updates |\n\nDexuan He 1,12, Xiao Zhou 2,12, Wenbin Guan3,12, Liyuan Zhang1, Xiaoman Zhang 4, Sinuo Xu1, Ge Wang 5, Lifeng Wang3, Xiaojun Yuan6, Jing Ma7, Xin Sun8, Yanfeng Wang 1, Kun Sun 9,10 , Ya Zhang 1,2,11  & Weidi Xie 1,2   \nRare cancers comprise 20–25% of malignancies (over 70% in pediatric oncology) but face major diagnostic challenges due to limited expert availability. While pathology vision-language models show promising zero-shot capabilities for common cancers, their performance on rare cancers remains limited. Existing multi-instance learning (MIL) methods rely solely on visual features, overlooking cross-modal knowledge and compromising interpretability critical for rare cancer diagnosis. To address this, we propose PathPT, a framework that exploits vision-language foundation models through spatiallyaware visual aggregation and task-speciﬁc prompt tuning. PathPT converts WSI-level supervision into ﬁne-grained tile-level guidance, preserving tumor localization and enabling cross-modal reasoning. Across eight rare and three common cancer datasets–spanning 56 subtypes and 3958 WSIs, PathPT consistently outperforms state-of-the-art methods under data-scarce settings. It achieves substantial gains in both subtyping accuracy and cancerous region grounding ability, providing a scalable, interpretable AI solution to improve rare cancer subtyping with limited access to specialized expertise.  \nRare cancers, while individually uncommon, collectively account for 20–25% of all malignancies and pose signiﬁcant diagnostic challenges due to a lack of clinical expertise and limited reference cases1. These challenges are particularly pronounced in pediatric oncology, where rare tumors comprise over 70% of diagnoses1, and timely, accurate subtyping is essential for effective treatment2,3. In many settings, for  \nexample, underserved regions and specialized clinics, the scarcity of experienced pathologists further underscores the urgent need for automated diagnostic tools capable of reliable performance under conditions of data scarcity and minimal supervision.  \nRecent advances in deep learning, particularly in selfsupervised4–7 and vision-language (VL)8–10 pretraining, have driven  \n1School of Artiﬁcial Intelligence, Shanghai Jiao Tong University, Shanghai, China. 2Shanghai Artiﬁcial Intelligence Laboratory, Shanghai, China. 3Department of Pathology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 4Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA. 5Department of Oral Pathology, Shanghai Ninth People’s Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 6Department of Pediatric Hematology/Oncology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 7Department of Pathology, Shanghai Children’s Medical Center, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 8Clinical Research and Innovation Unit, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 9Department of Pediatric Cardiology, Xinhua Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China. 10Engineering Research Centre of Techniques and Instruments for Diagnosis and Treatment of Congenital Heart Disease, Ministry of Education, Shanghai, China. 11Institute of Artiﬁcial Intelligence for Medicine, School of Medicine, Shanghai Jiao Tong University, Shanghai, China. 12These authors contributed equally: Dexuan He, Xiao Zhou, Wenbin Guan. e-mail: [sunkun@xinhuamed.com.cn](sunkun@xinhuamed.com.cn);  \n[ya_zhang@sjtu.edu.cn](ya_zhang@sjtu.edu.cn); [weidi@sjtu.edu.cn](weidi@sjtu.edu.cn)  \nthe development of pathology foundation m","cbCaigOWSQJpSv4X","https://ap.wps.com/l/cbCaigOWSQJpSv4X","pdf",3488106,13,"English","# Introduction\n## Background and challenges in rare cancer diagnosis\n## Pathology foundation models and vision-language pretraining\n# Method\n## Framework overview: PathPT\n## Spatially-aware visual aggregation\n## Task-adaptive prompt tuning\n# Experiments\n## Dataset coverage and evaluation under data-scarce settings\n## Subtyping accuracy and region grounding performance\n# Results and implications","[{\"question\":\"Why are rare cancer subtyping tasks challenging in practice?\",\"answer\":\"Rare cancers are collectively common but individually uncommon, and they lack clinical expertise and annotated reference cases. This limits reliable training signals and makes accurate, timely subtyping difficult.\"},{\"question\":\"What limitations exist in adapting vision-language foundation models using the current MIL paradigm?\",\"answer\":\"Existing MIL approaches focus mainly on visual features and ignore cross-modal reasoning from the textual encoder. They also rely on slide-level supervision, providing weak spatial guidance for fine-grained region-specific patterns.\"},{\"question\":\"How does PathPT improve rare cancer subtyping under few-shot or data-scarce conditions?\",\"answer\":\"PathPT introduces spatially-aware visual aggregation and task-specific prompt tuning to harness vision-language prior knowledge. It converts WSI-level supervision into fine-grained tile-level guidance, improving both subtyping accuracy and tumor region grounding.\"}]","Boosting pathology foundation models via few-shot prompt-tuning for rare cancer subtyping | PDF",1790053877,33,[148,152,156,160],{"page_no":22,"url":149,"width":150,"height":151},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/1fd83ab862984a0466cbc03faff95e46.webp",983,1305,{"page_no":26,"url":153,"width":154,"height":155},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/36bf67e2eccd13f1ce585e4c4799e234.webp",982,1304,{"page_no":30,"url":157,"width":158,"height":159},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/7d71b8812ba828d57b45df0a6618b7f7.webp",981,1303,{"page_no":33,"url":161,"width":162,"height":163},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/956d0b0a39b1ae47575a4b21aa4684b9.webp",980,1301]