[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-362356-105":59,"doc-detail-362356-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","foundation-model-embeddings-for-multimodal-oncology-data-integration","Foundation model embeddings for multimodal oncology data integration","","Cancer care generates extensive clinical, imaging, and molecular data, but routine workflows rarely integrate these sources systematically, leaving them fragmented across separate systems. Foundation model–driven embeddings provide a way to convert text, radiology, whole-slide pathology, and molecular profiles into unified vector representations for patient-level similarity and classification. The discussion centers on the HONeYBEE platform, its modular architecture, and key clinical and technical barriers that affect real-world implementation.",{"@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":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/foundation-model-embeddings-for-multimodal-oncology-data-integration/362356/",{"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/foundation-model-embeddings-for-multimodal-oncology-data-integration/362356.png","ImageObject",300,407,{"name":92,"@type":93},"Mia  ","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-23","2026-09-23",true,{"@type":102,"interactionType":103,"userInteractionCount":52},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why is integrating oncology data across modalities difficult in routine clinical workflows?","Question",{"text":112,"@type":113},"Different modalities (clinical text, pathology images, radiology scans, and molecular profiles) are often siloed across platforms and infrastructure, so they are rarely interpreted together meaningfully.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What does the HONeYBEE platform do for multimodal integration?",{"text":117,"@type":113},"It uses foundation model–based embeddings to generate fixed-length vectors from clinical text, pathological reports and images, radiology scans and whole-slide images, and molecular profiles to support cancer classification and patient similarity retrieval.",{"name":119,"@type":110,"acceptedAnswer":120},"What infrastructure challenges does HONeYBEE aim to address?",{"text":121,"@type":113},"It applies a modular design that standardizes preprocessing and embedding generation for each data type, helping reduce overhead from coordinating separate tools, formats, and preprocessing requirements.","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},362356,1790197962,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":52,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":19,"language":139,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":140,"faqs":141,"seo_title":142,"seo_description":67,"update_tm":143,"read_time":52,"preview_page_count":19,"preview_urls":144},687207024478,"https://ap-avatar.wpscdn.com/davatar_a8503ba1806abce46bf441b54a3ca4cd","npj | digital medicine News & Views  \n\n| Published in partnership with Seoul National University Bundang Hospital | |\n| --- | --- |\n| [https://doi.org/10.1038/s41746-025-02312-8](https://doi.org/10.1038/s41746-025-02312-8) |  |\n| Foundation model embeddings for multimodal oncology data integration\u003Cbr>Tara P. Menon, Arjun Mahajan & Dylan Powell  Check for updates |  |\n\nCancer care generates vast quantities of data including clinical records, pathology images, radiology scans, and molecular proﬁles, yet these modalities are rarely integrated in a systematic, automated manner within routine clinical workﬂows, remaining largely siloed across separate departmental and technical systems. Foundation model-driven embeddings—or numerical representations (vectors) that summarize complex data such as text, images, and molecular proﬁles —offer a framework to integrate these data streams into uniﬁed patient representations. Here we examine the HONeYBEE platform ’s approach to multimodal integration in oncology, situate it within broader developments in representation learning, and clinical and technical challenges that may shape its path to implementation1.  \nCancer care is increasingly driven and deﬁned by data—radiology, pathology, genomics, and electronic health records all capture different dimensions of the same disease, yet rarely connect meaningfully. The scale and fragmentation of this data have created both an opportunity and an unmet need: howto synthesize these modalities into auniﬁed representation of the patient. The contemporary cancer patient generates an evolving digital footprint—a dynamic record encompassing both structured data (cancer stage, performance status, and treatment history) and unstructured data (narrative patient notes capturing symptom evolution and clinical reasoning)2. A pathologist can review a gigapixel whole-slide image revealing a patient’s histological architecture and cellular detail, while a radiologist interprets volumetric CT and MRI scans tracking tumor burden and treatment response. Genomic sequencing further extends this digital footprint to the molecular drivers of disease3.  \nYet despite the coexistence in electronic health records, these modalities are rarely interpreted together. Radiologists often review images without detailed molecular context, and pathologists assess tissue samples with limited integration of corresponding imaging or longitudinal clinical data. This separation persists atthe digital level, where oncology data remainsiloed across platforms, formats, and infrastructure, limiting their collective potential in precision oncology.  \nThis fragmentation highlights a central unmet need in oncology: auniﬁed framework capable of harmonizing heterogeneous data into a shared analytical space.  \nA recent article in npj Digital Medicine, Tripathi et al. propose HONeYBEE (Harmonized Oncology Biomedical Embedding Encoder), an open-source platform that helps solve these integration challenges through foundation model-based embeddings1. The platform generates numerical representations (vectors) from clinical text, pathological reports, radiology images, whole-slide images, and molecular proﬁles. The article sheds light on multimodal integration as a feasible and predictive technique for cancer classiﬁcation and patient similarity retrieval without requiring a full dataset for each patient. However, it also raises questions about the platform’s applicability in real-world settings. Here we examine the HONeYBEE platform’s approach to multimodal integration in oncology, situate it within broader developmentsin representation learning, and clinical and technical challenges that may shape its path to implementation.  \nThe data fragmentation problem  \nThe need to integrate cancer datasets has been recognized for more than a decade. Early studies sought to link patient-level clinical outcomes with one type of complex dataset, such as gene expression proﬁles or pathology imaging, rather than com","cbCaivO5ztDKBObl","https://ap.wps.com/l/cbCaivO5ztDKBObl","pdf",1198065,"English","# The data fragmentation problem\n## HONeYBEE’s modular architecture","[{\"question\":\"Why is integrating oncology data across modalities difficult in routine clinical workflows?\",\"answer\":\"Different modalities (clinical text, pathology images, radiology scans, and molecular profiles) are often siloed across platforms and infrastructure, so they are rarely interpreted together meaningfully.\"},{\"question\":\"What does the HONeYBEE platform do for multimodal integration?\",\"answer\":\"It uses foundation model–based embeddings to generate fixed-length vectors from clinical text, pathological reports and images, radiology scans and whole-slide images, and molecular profiles to support cancer classification and patient similarity retrieval.\"},{\"question\":\"What infrastructure challenges does HONeYBEE aim to address?\",\"answer\":\"It applies a modular design that standardizes preprocessing and embedding generation for each data type, helping reduce overhead from coordinating separate tools, formats, and preprocessing requirements.\"}]","Foundation model embeddings for multimodal oncology data integration | PDF",1790147282,[145,149,153,157],{"page_no":8,"url":146,"width":147,"height":148},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/30ffd1f4386ee4cce41b36ec5b679916.webp",983,1305,{"page_no":14,"url":150,"width":151,"height":152},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/7134bd0e4ec6c293c3f8f8faaf4385d4.webp",982,1304,{"page_no":81,"url":154,"width":155,"height":156},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/5eac14e3487f6c110e98fb86e111ed63.webp",981,1303,{"page_no":19,"url":158,"width":159,"height":160},"https://res-academy.cache.wpscdn.com/upload/kdocshare/2026-11-16/ef6bc9d8006f18403110de490f137ae9.webp",980,1301]