[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85012-en":3,"doc-seo-85012-105":29,"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":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":13,"seo_description":14,"update_tm":27,"read_time":28},85012,13056703019404,"Miles","https://ap-avatar.wpscdn.com/davatar_29158cc5080c5b710cf443261637dec0",8,"Research & Report","MedPMC: A Systematic Framework for Scaling High-Fidelity Medical Multimodal Data for Foundation Models","Medicine relies on multimodal evidence, yet multimodal foundation models are limited by scarce access to large-scale, high-quality clinical data. PubMed Central (PMC) provides expert image-text literature, but existing PMC-derived resources fall short in fidelity, reproducibility, and clinical validation. MedPMC delivers an automated, continuously updatable framework that converts permissively licensed PMC literature into high-fidelity medical multimodal infrastructure. Trained on 6.1M PMC papers, it curates 11M image-text pairs and improves benchmark and clinical tasks.","arXiv :2607 .07673v 1 [ cs .CV] 8 Jul 2026  \nMedPMC: A Systematic Framework for Scaling High-Fidelity Medical Multimodal Data for Foundation Models  \nHyunjae Kim 1 , Dain Kim2 , Pan Xiao3 , Serina S. Applebaum 1 , Younjoon Chung 1 , Xuguang Ai 1 , Yu Yin4 , Roy Jiang 1 , Yuexi Du 1 , Yawen Wei 1 , Yiming Kong 1 , Tuo Guo 1 , Zhiyuan Cao 1 , Mengmeng Du 1 , Yuelei Fu 1 , Yan Hu5 , Rui Shi 1 , Gui Yang 1 , Kevin W. Jin 1 , Yuntian Liu 1 , Yuxuan Tian 1 , Jonathan Marquez6 , Zhen Chen 1 , Sheng Zhang7 , Hoifung Poon7 , Hua Xu 1 , Jaewoo Kang2 , and Qingyu Chen 1,*  \n1Yale University, New Haven, CT, USA  \n2 Korea University, Seoul, South Korea  \n3Washington University in St. Louis, St. Louis, MO, USA  \n4The University of Queensland, Brisbane, QLD, Australia  \n5The University of Texas Health Science Center at Houston, Houston, TX, USA  \n6 University of Washington, Seattle, WA, USA  \n7 Microsoft Research, Redmond, WA, USA  \n* [qingyu.chen@yale.edu](qingyu.chen@yale.edu)  \nABSTRACT  \nMedicine is inherently multimodal, requiring clinicians to synthesize information across diverse data streams. Yet the development of multimodal foundation models is constrained by limited access to large-scale, high-quality clinical data. Although PubMed Central (PMC) offers a complementary source of expert-authored image-text data, existing PMC-derived resources remain limited in fidelity, reproducibility, and clinical validation. We introduce MedPMC, an automated, continuously updatable framework that transforms permissively licensed literature into high-fidelity infrastructure for medical multimodal models. Applied to 6.1 million PMC articles, MedPMC curated 11 million medical image-text pairs. Component evaluations showed strong performance for initial screening (F1 = 93.2), multi-panel figure detection (F1 = 96.5), figure separation (mAP = 89.8), caption separation and alignment (F1 = 81 .4; ROUGE-L = 85.3), and medical figure classification (F1 = 96.5) . Manual review by five annotators, three with medical training, found 95.3% of MedPMC images medically relevant, versus 19.7% in a prior PMC-derived dataset. Across 26 benchmarks spanning 11 specialties, a MedPMC-trained CLIP-style model improved average zero-shot AUC by 7.1 percentage points over the strongest architecture-matched biomedical CLIP baseline despite using fewer than half as many image-text pairs. As the vision encoder in a multimodal large language model, it improved medical visual question-answering by 1 .9 and 16.9 percentage points across two benchmarks. In 10,524 Yale New Haven Health System dermatology photographs, it improved morphology-to-image retrieval Recall@5 by 11 .7 percentage points. These findings show that high-fidelity literature curation strengthens medical multimodal foundation models across benchmark and clinical settings. We publicly release the framework, corpus, benchmarks, and pretrained models.  \n1 Introduction  \nMedicine is multimodal in nature 1, 2. Triage, diagnosis, prognosis, and treatment planning routinely require clinicians to integrate images, text, laboratory values, and other complementary data streams3, 4. In principle, medical foundation models could unlock this potential by providing reusable backbones that generalize across medical modalities and can be adapted to local clinical settings with limited labeled  \ndata5, 6. In practice, however, current medical AI systems are still largely developed and evaluated within single-modality, single-task settings rather than on the heterogeneous evidence used in clinical care7, 8. This limitation stems not from model architecture alone, but from the absence of large-scale, high-quality multimodal data that are publicly accessible for training and evaluation, shareable across institutions, and suitable for reproducible research7–10. Unlike curated benchmark datasets, real-world medical data are fragmented across devices and health systems; constrained by privacy, licensing, and governance requirements; an","cbCaieoEfi2u3lnc","https://ap.wps.com/l/cbCaieoEfi2u3lnc","pdf",2249746,1,32,"English","en",105,"# Abstract\n# Introduction","[{\"question\":\"What problem does MedPMC address for medical multimodal foundation models?\",\"answer\":\"MedPMC targets the lack of large-scale, high-quality, publicly accessible multimodal clinical data. It also addresses shortcomings of existing PMC-derived resources in fidelity, reproducibility, and clinical validation.\"},{\"question\":\"How does MedPMC build its high-fidelity medical multimodal dataset?\",\"answer\":\"MedPMC provides an automated framework that transforms permissively licensed biomedical literature from PubMed Central into curated medical image-text pairs. Applied to 6.1 million PMC articles, it produces about 11 million pairs.\"},{\"question\":\"What performance improvements does MedPMC report across tasks and benchmarks?\",\"answer\":\"MedPMC-trained models improve average zero-shot AUC by 7.1 percentage points over a matched biomedical CLIP baseline, while using fewer than half the image-text pairs. It also boosts medical visual question-answering and morphology-to-image retrieval performance in reported benchmarks.\"}]",1784200249,81,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"medpmc-a-systematic-framework-for-scaling-high-fidelity-medical-multimodal-data-for-foundation-models","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/medpmc-a-systematic-framework-for-scaling-high-fidelity-medical-multimodal-data-for-foundation-models/85012/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","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},"What problem does MedPMC address for medical multimodal foundation models?","Question",{"text":75,"@type":76},"MedPMC targets the lack of large-scale, high-quality, publicly accessible multimodal clinical data. It also addresses shortcomings of existing PMC-derived resources in fidelity, reproducibility, and clinical validation.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does MedPMC build its high-fidelity medical multimodal dataset?",{"text":80,"@type":76},"MedPMC provides an automated framework that transforms permissively licensed biomedical literature from PubMed Central into curated medical image-text pairs. Applied to 6.1 million PMC articles, it produces about 11 million pairs.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance improvements does MedPMC report across tasks and benchmarks?",{"text":84,"@type":76},"MedPMC-trained models improve average zero-shot AUC by 7.1 percentage points over a matched biomedical CLIP baseline, while using fewer than half the image-text pairs. It also boosts medical visual question-answering and morphology-to-image retrieval performance in reported benchmarks.","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":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":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]