[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-125375-en":3,"doc-seo-125375-105":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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},125375,13056703019662,"Evangeline","https://ap-avatar.wpscdn.com/avatar/be000253a8e92610077?_k=1778726343310543188",8,"Research & Report","Latent Diffusion Models with Image-Derived Annotations for Enhanced AI-Assisted Cancer Diagnosis in Histopathology","Artificial intelligence image analysis holds major promise for supporting diagnostic histopathology, yet supervised learning typically depends on large-scale annotated datasets that are difficult to obtain. This work proposes a data augmentation approach using latent diffusion models to generate synthetic images and constructs structured textual prompts from automatically extracted image features. Experiments on the PCam dataset show improved synthetic quality (FID 178.8 to 90.2) and that pathologists can find synthetic images challenging (median sensitivity/specificity 0.55/0.55).","15 Dec 2023  \nLatent Diffusion Models with Image-Derived Annotations for Enhanced AI-Assisted Cancer Diagnosis in Histopathology  \nPedro Osorio Guillermo Jimenez-Perez Javier Montalt-Tordera Jens Hooge Guillem Duran-Ballester Shivam Singh Moritz Radbruch Ute Bach Sabrina Schroeder  \nKrystyna Siudak Julia Vienenkoetter Bettina Lawrenz Sadegh Mohammadi 1  \nBayer AG  \n[1](1 sadegh. mohammadi@bayer. com)[ sadegh. mohammadi@bayer. com](1 sadegh. mohammadi@bayer. com)  \nAbstract  \nArtificial Intelligence (AI) based image analysis has an immense potential to support diagnostic histopathology, including cancer diagnostics. However, developing supervised AI methods requires large-scale annotated datasets. A potentially powerful solution is to augment training data with synthetic data. Latent diffusion models, which can generate high-quality, diverse synthetic images, are promising. However, the most common implementations rely on detailed textual descriptions, which are not generally available in this domain. This work proposes a method that constructs structured textual prompts from automatically extracted image features. We experiment with the PCam dataset, composed of tissue patches only loosely annotated as healthy or cancerous. We show that including image-derived features in the prompt, as opposed to only healthy and cancerous labels, improves the Fr´echet Inception Distance (FID) from 178.8 to 90.2 . We also show that pathologists find it challenging to detect synthetic images, with a median sensitivity/specificity of 0.55/0.55 . Finally, we show that synthetic  \narXiv :23 12 .09792v1  \nvolves the microscopic examination of tissue samples to discern manifestations of disease. These tissue samples, typically obtained through surgical resections or biopsies, are prepared and analyzed using hematoxylin and eosin (H&E) staining protocols and can be digitized into gigapixel-sized whole-slide images (WSI) . H&E imaging offers insights into structural and morphological changes associated with various pathological conditions, including cancer. Due to their cost-effectiveness and accessibility, AI-based computer-assisted diagnosis (CAD) has already demonstrated immense potential by classifying diseases, detecting genetic alterations or quantifying lesions [1–4] .  \nHowever, applying this technology in the medical field, is challenging. First, collecting large enough datasets for model training is challenging due to disease rarity, to high acquisition costs and to reliance on low-availability technologies such as next-generation sequencing [1, 3] . Second, histopathologyspecific challenges arise, given the sheer size of the gigapixelsized WSIs and the high variability of different staining and slide preparation techniques. This further complicates matters, as AI models often struggle to generalize, even with intricate model architectures [1, 3, 5] . While traditional data augmentation offers a potential solution to address some of these issues, methods such as flipping and cropping often cannot adequately cover the full data distribution, whereas in-domain data augmentation techniques such as stain normalization can only slightly improve model generalization [3, 5] . This results in suboptimal improvements, as these transformations cannot effectively bridge the gaps introduced by missing data samples or address data imbalance or bias [6] .  \nGenerative models are pivotal tools in the image synthesis field and have been used for many applications, ranging from bias mitigation, by augmenting an underrepresented class in adataset, to privacy preservation, producing images that aren’t derived from real subjects [7, 8] . While some models, such as generative adversarial networks (GANs) have shown tremendous potential in generating diverse, distribution-wide and high-fidelity images, [9–12], these often struggle with mode collapse and training instability, which hinders model training and demands meticulous hyperparameter adjustments [13] .  \nL","cbCaikicsYj4we5Y","https://ap.wps.com/l/cbCaikicsYj4we5Y","pdf",8127069,1,17,"English","en",105,"# Abstract\n## Problem: limited annotated histopathology data\n## Proposed method: image-feature prompt construction for LDM\n## Experiments: PCam dataset and evaluation metrics\n## Human evaluation: pathologist detection of synthetic images","[{\"question\":\"Why are large annotated datasets a bottleneck for supervised AI in histopathology?\",\"answer\":\"Supervised methods require large-scale annotations, but histopathology data is hard to collect due to disease rarity, acquisition costs, and limited availability of enabling technologies.\"},{\"question\":\"What is the key idea of the proposed latent diffusion approach?\",\"answer\":\"The method generates synthetic histopathology images with latent diffusion and builds structured textual prompts from automatically extracted image features rather than relying only on basic labels.\"},{\"question\":\"How do the authors evaluate whether the synthetic images are useful and detectable?\",\"answer\":\"They assess synthetic quality using FID improvements on the PCam dataset and also perform a pathologist study to measure sensitivity/specificity when detecting synthetic images.\"}]","Latent Diffusion Models with Image-Derived Annotations for Enhanced AI-Assisted Cancer Diagnosis in Histopathology | PDF",1785898536,43,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"latent-diffusion-models-with-image-derived-annotations-for-enhanced-ai-assisted-cancer-diagnosis-in-histopathology","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"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":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/latent-diffusion-models-with-image-derived-annotations-for-enhanced-ai-assisted-cancer-diagnosis-in-histopathology/125375/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-05",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"Why are large annotated datasets a bottleneck for supervised AI in histopathology?","Question",{"text":75,"@type":76},"Supervised methods require large-scale annotations, but histopathology data is hard to collect due to disease rarity, acquisition costs, and limited availability of enabling technologies.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the key idea of the proposed latent diffusion approach?",{"text":80,"@type":76},"The method generates synthetic histopathology images with latent diffusion and builds structured textual prompts from automatically extracted image features rather than relying only on basic labels.",{"name":82,"@type":73,"acceptedAnswer":83},"How do the authors evaluate whether the synthetic images are useful and detectable?",{"text":84,"@type":76},"They assess synthetic quality using FID improvements on the PCam dataset and also perform a pathologist study to measure sensitivity/specificity when detecting synthetic images.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"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":53,"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"]