[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-346352-105":59,"doc-detail-346352-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","geometric-multi-instance-learning-for-weakly-supervised-gastric-cancer-segmentation","Geometric multi-instance learning for weakly supervised gastric cancer segmentation","","Weakly supervised segmentation of cancerous regions in whole-slide images (WSIs) is vital in computational pathology but limited by the high cost of pixel-level annotations. Multiple Instance Learning often treats WSIs as unordered bag-of-patches, losing tissue topology and architectural context. This work proposes Geometric Multi-Instance Learning (Geo-MIL), a graph-based method with a topological attention mechanism that models spatial relationships between tissue patches. Experiments on three public gastric cancer datasets show Geo-MIL substantially outperforms state-of-the-art baselines for segmentation and classification, bridging slide-level labels to pixel-level predictions.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":35,"@type":76,"position":81},"https://docshare.wps.com/document/healthcare/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/geometric-multi-instance-learning-for-weakly-supervised-gastric-cancer-segmentation/346352/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/geometric-multi-instance-learning-for-weakly-supervised-gastric-cancer-segmentation/346352.png","ImageObject",300,407,{"name":92,"@type":93},"Himbo","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-28","2026-09-22",true,{"@type":102,"interactionType":103,"userInteractionCount":44},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why is pixel-level annotation a major obstacle for gastric cancer WSI segmentation?","Question",{"text":112,"@type":113},"Accurate tumor boundary masks require labor-intensive, time-consuming work by expert pathologists. This creates a scalability bottleneck for training robust segmentation models.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What limitation of conventional MIL harms segmentation mask accuracy?",{"text":117,"@type":113},"Conventional MIL assumes instances in a bag are independent and identically distributed, ignoring spatial context and tissue architecture critical for malignancy localization.",{"name":119,"@type":110,"acceptedAnswer":120},"How does Geo-MIL address the bag-of-patches limitation?",{"text":121,"@type":113},"Geo-MIL represents a WSI as a graph of patches with spatial adjacency edges and uses a learnable topological gating mechanism plus topological attention to capture diagnostically relevant tissue structures.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},346352,1790207630,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":34,"category_name":35,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":44,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145},687197100911,"https://ap-avatar.wpscdn.com/avatar/a000239b6f1da00475?x-image-process=image/resize,m_fixed,w_180,h_180&k=1785132997149421697","npj | digital medicine Article  \n\n| Published in partnership with Seoul National University Bundang Hospital |  | |\n| --- | --- | --- |\n| [https://doi.org/10.1038/s41746-025-02287-6](https://doi.org/10.1038/s41746-025-02287-6) |  |  |\n| Geometric multi-instance learning for weakly supervised gastric cancer segmentation\u003Cbr> Check for updates |  |  |\n| Chenshen Huang1,2,6 , Haoyun Xia1,6, Xi Xiao3,6, Hong Chen1,4, Yiqing Jiang5, Yahui Lyu2, Zhizhan Ni4, Tianyang Wang3 , Ning Wang2  & Qi Huang4  |  |  |\n| Weakly supervised segmentation of cancerous regions in whole-slide images(WSIs)is acrucial task in computational pathology, but it is severely hampered by the need for expensive pixel-level annotations. Existing Multiple Instance Learning (MIL) frameworks, while popular, typically fail to produce accurate segmentation masks because they treat WSIs as an unordered ’bag-of-patches’, ignoring the critical tissue topology and architectural patterns that define malignancy. In this paper, we address this fundamental limitation by proposing Geometric Multi-Instance Learning (Geo-MIL), a novel graph-based framework that explicitly models the spatial relationships between tissue patches. At the core of our method is a new topological attention mechanism that operates on the WSI graph, learning to identify and prioritize entire diagnostically relevant tissue structures over isolated patch features. Through extensive experiments on three public gastric cancer datasets, we demonstrate that Geo-MIL significantly outperforms a wide array of state-of-the-art baselines, achieving a new benchmark in both segmentation accuracy and classification performance. Our work represents asignificant step towards bridging the gap between weak slide-level labels and precise, pixel-level predictions, paving the way for scalable and accurate quantitative analysis in digital pathology. |  |  |\n| Gastric cancer (GC) is a leading cause of cancer-related mortality worldwide, with a significant disease burden1. The gold standard for its diagnosis and staging relies on the histopathological assessment oftissue biopsies. This process involves pathologists meticulously examining whole-slide images (WSIs) to identify malignant regions, assess morphological features, and determine the extent of tumor invasion. Recently, deep learning models have shown remarkable promise in automating parts of this workflow, offering the potential to improve diagnostic efficiency and reproducibility2.\u003Cbr>A primary obstacle to developing highly accurate deep learning models for GC segmentation is the immense data annotation requirement. Fully supervised methods, such as U-Net and its variants, depend on large datasets with precise, pixel-level annotations delineating tumor boundaries. Creating these detailed masks is exceptionally laborious and timeconsuming. It demands the focused effort of expert pathologists, making the process asignificant bottleneck that hinders the development ofscalable and robust models3. | To circumvent this annotation bottleneck, Weakly Supervised Learning(WSL)has gained significant traction. Multiple Instance Learning(MIL) stands out asthe dominant WSL paradigm for computational pathology4. In theMIL framework, aWSI is treated asa “bag\"ofimage patches(instances), and the model is trained using only a single, slide-level label(e.g.,“tumor\"or“normal\"). Attention-based MIL models can successfully identify discriminative, tumor-bearing patches and have achieved state-of-the-art results in WSI classification tasks5.\u003Cbr>However, the conventional MIL framework suffers from a fundamental limitation: it operates on the assumption that instances within a bag are independent and identically distributed. This “bag-of-patches\"approach disregards the critical spatial context and tissue architecture inherent in histopathology. The growth patterns of gastric cancer, such as glandular formation, stromal invasion, and cell differentiation, are defined by the spatial relationships between ","cbCaipkTKhWskGWa","https://ap.wps.com/l/cbCaipkTKhWskGWa","pdf",2395536,14,"English","# Introduction\n## Clinical motivation and annotation bottleneck\n## Limits of conventional MIL (bag-of-patches)\n# Proposed method\n## Geo-MIL graph modeling of WSIs\n## Topological attention and gating mechanism\n# Contributions and evaluation","[{\"question\":\"Why is pixel-level annotation a major obstacle for gastric cancer WSI segmentation?\",\"answer\":\"Accurate tumor boundary masks require labor-intensive, time-consuming work by expert pathologists. This creates a scalability bottleneck for training robust segmentation models.\"},{\"question\":\"What limitation of conventional MIL harms segmentation mask accuracy?\",\"answer\":\"Conventional MIL assumes instances in a bag are independent and identically distributed, ignoring spatial context and tissue architecture critical for malignancy localization.\"},{\"question\":\"How does Geo-MIL address the bag-of-patches limitation?\",\"answer\":\"Geo-MIL represents a WSI as a graph of patches with spatial adjacency edges and uses a learnable topological gating mechanism plus topological attention to capture diagnostically relevant tissue structures.\"}]","Geometric multi-instance learning for weakly supervised gastric cancer segmentation | PDF",1790061024,35]