[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85148-en":3,"doc-seo-85148-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},85148,1374391974468,"Eden","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","Slide-Level Active Learning Reduces Annotation Burden in H&E Images","Deep learning semantic segmentation of histopathology whole-slide images (WSIs) depends on expensive pixel-level annotations. Existing active learning (AL) approaches fall short because uncertainty estimates fail on partially annotated WSIs, patch-level acquisition mismatches slide-level annotation workflows, and multi-class class imbalance is not explicitly handled. SHAL (Slide-level Hybrid Active Learning) is introduced as a patient-level AL framework for annotation-efficient multi-class histopathology segmentation. SHAL suppresses unlabeled background bias, combines entropy with epistemic uncertainty stage-adaptively, and prioritizes diagnostically relevant tissue classes, evaluated on TCGA colorectal cancer data.","arXiv :2607 .09831v1 [ ee ss .IV] 10 Jul 2026  \nSlide-Level Active Learning Reduces Annotation Burden in H&E images  \nMahsa Valia,b,d,∗, Zhilong Wenge , Noémie Moreaua,b , Yuri Tolkache ,  \nKatarzyna Bozeka,b,c  \na Institute for Biomedical Informatics, Faculty of Medicine and University Hospital  \nCologne, University of Cologne, Cologne, Germany  \nb Center for Molecular Medicine Cologne (CMMC), Faculty of Medicine and University  \nHospital Cologne, University of Cologne, Cologne, Germany  \nc Cologne Excellence Cluster on Cellular Stress Responses in Aging-Associated Diseases  \n(CECAD), University of Cologne, Cologne, Germany  \nd Faculty of Mathematics and Natural Sciences, University of Cologne, Cologne, Germany e Institute of Pathology, University Hospital Cologne, Cologne, Germany  \nAbstract  \nDeep learning-based segmentation of histopathology whole-slide images (WSIs) requires large amounts of pixel-level annotations, which are costly and timeconsuming to obtain. Active learning (AL) has been proposed to reduce this effort, but existing methods exhibit three key limitations. Uncertainty estimation is unreliable on partially annotated WSIs, patch-level acquisition is inconsistent with slide-level annotation workflows, and class imbalance in multi-class settings is not explicitly addressed. To address these challenges, propose SHAL (Slide-level Hybrid Active Learning), a patient-level AL framework, is introduced for annotation-efficient multi-class histopathology segmentation. SHAL integrates three complementary components. A foreground-aware strategy suppresses bias from unlabeled background regions. A stage-adaptive mechanism hybridizes predictive entropy and epistemic uncertainty across learning stages. A class-aware strategy prioritizes diagnostically relevant tissue classes. SHAL is evaluated on the TCGA colorectal cancer dataset. It achieves the highest Macro Dice at the full annotation budget (0.846) and reaches Dice ≥ 0.80 using only 26% of the budget (50 of 190 slides), whereas competing methods reach this threshold only at  \n∗ Corresponding author  \nEmail address: [mahsavali14@gmail.com](mahsavali14@gmail.com) (Mahsa Vali)  \n37%(70 slides) . Across five independent external cohorts, SHAL attains the highest mean external Macro Dice FG (0.815) and the smallest internal-toexternal generalization gap among all methods (0 .025 at Round 3, 0.026 atthe full budget) . The results indicate that patient-level hybrid uncertainty acquisition reduces annotation cost without sacrificing cross-domain generalization in computational pathology.  \nKeywords:  \nActive Learning, Annotation Efficiency, Whole Slide Image (WSI), Semantic Segmentation, Uncertainty-Based Sampling  \n1. Introduction  \nComputational pathology increasingly integrates artificial intelligence (AI) into digital pathology workflows, enabling quantitative analysis of tissue morphology and supporting clinical decision-making [1, 2] . A central task in this area is semantic segmentation, which provides pixel-level delineation of tissue components such as tumor, stroma, necrosis, lymphocytes, and other structures. Accurate characterization of these regions is essential for understanding the tumor microenvironment, assessing disease progression, and developing robust prognostic biomarkers [3] . However, producing dense pixellevel annotations for whole-slide images (WSIs) remains a major bottleneck in computational pathology [4] . WSIs are extremely large, often reaching gigapixel resolution, and their annotation requires extensive work by expert pathologists [5] . This process is time-consuming, expensive, and subject to inter-observer variability [6] . Unlike natural image datasets, annotation in histopathology requires specialized expertise and cannot be easily crowdsourced [7] . As a result, annotation cost significantly limits the scalability of AI-based pathology systems.  \nDeep learning has substantially advanced histopathology segmentation. Convolutional encoder–","cbCaigOYZQj9UwLT","https://ap.wps.com/l/cbCaigOYZQj9UwLT","pdf",4172135,3,1,26,"English","en",105,"# Introduction\n## Computational Pathology and Segmentation\n## Bottleneck of Pixel-Level Annotation\n## Active Learning for Medical Image Segmentation\n## Limitations of Existing Histopathology AL Methods\n## Motivation for SHAL","[{\"question\":\"Why is annotation a bottleneck in histopathology whole-slide image segmentation?\",\"answer\":\"WSIs are extremely large (often gigapixel scale), requiring extensive expert pixel-level work. This makes annotation time-consuming, costly, and vulnerable to inter-observer variability.\"},{\"question\":\"What limitations do existing active learning methods have for histopathology segmentation?\",\"answer\":\"Uncertainty estimation is unreliable on partially annotated WSIs, patch/region acquisition does not align well with slide-level workflows, and multi-class class imbalance is not explicitly addressed.\"},{\"question\":\"How does SHAL reduce annotation cost while keeping generalization strong?\",\"answer\":\"SHAL uses a foreground-aware strategy to avoid bias from unlabeled background, a stage-adaptive hybrid of predictive entropy and epistemic uncertainty, and a class-aware selection that prioritizes diagnostically relevant tissue classes. Experiments on TCGA and external cohorts show reduced budget usage with high external Macro Dice and small generalization gaps.\"}]",1784201390,66,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"slide-level-active-learning-reduces-annotation-burden-in-he-images","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/slide-level-active-learning-reduces-annotation-burden-in-he-images/85148/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-24","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why is annotation a bottleneck in histopathology whole-slide image segmentation?","Question",{"text":75,"@type":76},"WSIs are extremely large (often gigapixel scale), requiring extensive expert pixel-level work. This makes annotation time-consuming, costly, and vulnerable to inter-observer variability.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What limitations do existing active learning methods have for histopathology segmentation?",{"text":80,"@type":76},"Uncertainty estimation is unreliable on partially annotated WSIs, patch/region acquisition does not align well with slide-level workflows, and multi-class class imbalance is not explicitly addressed.",{"name":82,"@type":73,"acceptedAnswer":83},"How does SHAL reduce annotation cost while keeping generalization strong?",{"text":84,"@type":76},"SHAL uses a foreground-aware strategy to avoid bias from unlabeled background, a stage-adaptive hybrid of predictive entropy and epistemic uncertainty, and a class-aware selection that prioritizes diagnostically relevant tissue classes. Experiments on TCGA and external cohorts show reduced budget usage with high external Macro Dice and small generalization gaps.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]