{"id":1179141,"date":"2026-07-20T06:41:50","date_gmt":"2026-07-20T13:41:50","guid":{"rendered":"https:\/\/research.codeghost.online\/en-us\/research\/publication\/medpmc-a-systematic-framework-for-scaling-high-fidelity-medical-multimodal-data-for-foundation-models\/"},"modified":"2026-07-21T14:41:35","modified_gmt":"2026-07-21T21:41:35","slug":"medpmc-a-systematic-framework-for-scaling-high-fidelity-medical-multimodal-data-for-foundation-models","status":"publish","type":"msr-research-item","link":"https:\/\/research.codeghost.online\/en-us\/research\/publication\/medpmc-a-systematic-framework-for-scaling-high-fidelity-medical-multimodal-data-for-foundation-models\/","title":{"rendered":"MedPMC: A Systematic Framework for Scaling High-Fidelity Medical Multimodal Data for Foundation Models"},"content":{"rendered":"\n\n\n<p class=\"wp-block-paragraph\">Medicine 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.<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Medicine 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 [&hellip;]<\/p>\n","protected":false},"featured_media":0,"template":"","meta":{"msr-url-field":"","msr-podcast-episode":"","msrModifiedDate":"","msrModifiedDateEnabled":false,"ep_exclude_from_search":false,"_classifai_error":"","msr-author-ordering":[{"type":"text","value":"Hyunjae Kim","user_id":0},{"type":"text","value":"Dain Kim","user_id":0},{"type":"text","value":"Pan Xiao","user_id":0},{"type":"text","value":"Serina S. Applebaum","user_id":0},{"type":"text","value":"Younjoon Chung","user_id":0},{"type":"text","value":"Xuguang Ai","user_id":0},{"type":"text","value":"Yu Yin","user_id":0},{"type":"text","value":"Roy Jiang","user_id":0},{"type":"text","value":"Yuexi Du","user_id":0},{"type":"text","value":"Yawen Wei","user_id":0},{"type":"text","value":"Yiming Kong","user_id":0},{"type":"text","value":"Tuo Guo","user_id":0},{"type":"text","value":"Zhiyuan Cao","user_id":0},{"type":"text","value":"Mengmeng Du","user_id":0},{"type":"text","value":"Yuelei Fu","user_id":0},{"type":"text","value":"Yan Hu","user_id":0},{"type":"text","value":"Ruihe Shi","user_id":0},{"type":"text","value":"Guimiao Yang","user_id":0},{"type":"text","value":"Kevin W. Jin","user_id":0},{"type":"text","value":"Yuntian Liu","user_id":0},{"type":"text","value":"Yu Tian","user_id":0},{"type":"text","value":"Jonathan Marquez","user_id":0},{"type":"text","value":"Zhen Chen","user_id":0},{"type":"user_nicename","value":"Sheng Zhang","user_id":"39087"},{"type":"user_nicename","value":"Hoifung Poon","user_id":"32016"},{"type":"text","value":"Hua Xu","user_id":0},{"type":"text","value":"Jaewoo Kang","user_id":0},{"type":"text","value":"Qingyu Chen","user_id":0}],"msr_publishername":"","msr_publisher_other":"","msr_booktitle":"","msr_chapter":"","msr_edition":"","msr_editors":"","msr_how_published":"arXiv","msr_isbn":"","msr_issue":"","msr_journal":"","msr_number":"","msr_organization":"","msr_pages_string":"","msr_page_range_start":"","msr_page_range_end":"","msr_series":"","msr_volume":"","msr_copyright":"","msr_conference_name":"","msr_doi":"","msr_arxiv_id":"2607.07673","msr_mag_id":"","msr_other_authors":"","msr_other_contributors":"","msr_speaker":"","msr_award":"","msr_affiliation":"","msr_institution":"","msr_host":"","msr_version":"","msr_duration":"","msr_release_tracker_id":"","msr_highlight_type":"","msr_date_display_format":"","msr_main_download_label":"","msr_external_link_label":"","msr_doi_label":"","msr_published_date":"2026-07-08","msr_startdate":"","msr_presentation_date":"","msr_highlight_text":"","msr_notes":"","msr_longbiography":"","msr_publicationurl":"","msr_external_url":"","msr_secondary_video_url":"","msr_conference_url":"","msr_journal_url":"","msr_year":2026,"msr_month":7,"msr_day":8,"msr_microsoftintellectualproperty":false,"msr_pub_id":"4978524d94c604aba23791d986027f31fc566973","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/abs\/2607.07673","label_id":243109,"id":false,"viewUrl":false}],"msr_related_uploader":[],"msr_original_fields_of_study":[],"msr_s2_paper_id":"4978524d94c604aba23791d986027f31fc566973","msr_s2_pdf_url":"","msr_citation_count_updated":"","msr_citation_count":0,"msr_influential_citations":0,"msr_reference_count":107,"msr_s2_open_access":false,"msr_s2_author_ids":[],"msr_pub_ids":[{"provider":"s2","id":"4978524d94c604aba23791d986027f31fc566973"},{"provider":"arxiv","id":"2607.07673"},{"provider":"corpusid","id":"289978719"}],"msr_hide_image_in_river":0,"footnotes":""},"msr-research-highlight":[],"research-area":[13556,13553],"msr-publication-type":[270373],"msr-publisher":[],"msr-publication-cta":[],"msr-focus-area":[],"msr-locale":[268875],"msr-post-option":[],"msr-field-of-study":[246691,246985],"msr-conference":[],"msr-journal":[],"msr-impact-theme":[],"msr-pillar":[],"class_list":["post-1179141","msr-research-item","type-msr-research-item","status-publish","hentry","msr-research-area-artificial-intelligence","msr-research-area-medical-health-genomics","msr-locale-en_us","msr-field-of-study-computer-science","msr-field-of-study-medicine"],"msr_publishername":"","msr_edition":"","msr_affiliation":"","msr_published_date":"2026-07-08","msr_host":"","msr_duration":"","msr_version":"","msr_speaker":"","msr_other_contributors":"","msr_booktitle":"","msr_pages_string":"","msr_chapter":"","msr_isbn":"","msr_journal":"","msr_volume":"","msr_number":"","msr_editors":"","msr_series":"","msr_issue":"","msr_organization":"","msr_how_published":"arXiv","msr_notes":"","msr_highlight_text":"","msr_release_tracker_id":"","msr_original_fields_of_study":"","msr_download_urls":"","msr_external_url":"","msr_secondary_video_url":"","msr_longbiography":"","msr_microsoftintellectualproperty":0,"msr_main_download":"","msr_publicationurl":"","msr_doi":"","msr_publication_uploader":[{"type":"url","title":"https:\/\/arxiv.org\/abs\/2607.07673","label_id":243109,"id":false,"viewUrl":false}],"msr_related_uploader":[],"msr_citation_count":0,"msr_citation_count_updated":"","msr_s2_paper_id":"4978524d94c604aba23791d986027f31fc566973","msr_influential_citations":0,"msr_reference_count":107,"msr_arxiv_id":"2607.07673","msr_s2_author_ids":[],"msr_s2_open_access":false,"msr_s2_pdf_url":null,"msr_attachments":[],"msr-author-ordering":[{"type":"text","value":"Hyunjae Kim","user_id":0,"rest_url":false},{"type":"text","value":"Dain Kim","user_id":0,"rest_url":false},{"type":"text","value":"Pan Xiao","user_id":0,"rest_url":false},{"type":"text","value":"Serina S. Applebaum","user_id":0,"rest_url":false},{"type":"text","value":"Younjoon Chung","user_id":0,"rest_url":false},{"type":"text","value":"Xuguang Ai","user_id":0,"rest_url":false},{"type":"text","value":"Yu Yin","user_id":0,"rest_url":false},{"type":"text","value":"Roy Jiang","user_id":0,"rest_url":false},{"type":"text","value":"Yuexi Du","user_id":0,"rest_url":false},{"type":"text","value":"Yawen Wei","user_id":0,"rest_url":false},{"type":"text","value":"Yiming Kong","user_id":0,"rest_url":false},{"type":"text","value":"Tuo Guo","user_id":0,"rest_url":false},{"type":"text","value":"Zhiyuan Cao","user_id":0,"rest_url":false},{"type":"text","value":"Mengmeng Du","user_id":0,"rest_url":false},{"type":"text","value":"Yuelei Fu","user_id":0,"rest_url":false},{"type":"text","value":"Yan Hu","user_id":0,"rest_url":false},{"type":"text","value":"Ruihe Shi","user_id":0,"rest_url":false},{"type":"text","value":"Guimiao Yang","user_id":0,"rest_url":false},{"type":"text","value":"Kevin W. Jin","user_id":0,"rest_url":false},{"type":"text","value":"Yuntian Liu","user_id":0,"rest_url":false},{"type":"text","value":"Yu Tian","user_id":0,"rest_url":false},{"type":"text","value":"Jonathan Marquez","user_id":0,"rest_url":false},{"type":"text","value":"Zhen Chen","user_id":0,"rest_url":false},{"type":"user_nicename","value":"Sheng Zhang","user_id":39087,"rest_url":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Sheng Zhang"},{"type":"user_nicename","value":"Hoifung Poon","user_id":32016,"rest_url":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/microsoft-research\/v1\/researchers?person=Hoifung Poon"},{"type":"text","value":"Hua Xu","user_id":0,"rest_url":false},{"type":"text","value":"Jaewoo Kang","user_id":0,"rest_url":false},{"type":"text","value":"Qingyu Chen","user_id":0,"rest_url":false}],"msr_impact_theme":[],"msr_research_lab":[849856],"msr_event":[],"msr_group":[],"msr_project":[],"publication":[],"video":[],"msr-tool":[],"msr_publication_type":"misc","related_content":[],"_links":{"self":[{"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1179141","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item"}],"about":[{"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/types\/msr-research-item"}],"version-history":[{"count":2,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1179141\/revisions"}],"predecessor-version":[{"id":1179471,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-item\/1179141\/revisions\/1179471"}],"wp:attachment":[{"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/media?parent=1179141"}],"wp:term":[{"taxonomy":"msr-research-highlight","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-research-highlight?post=1179141"},{"taxonomy":"msr-research-area","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/research-area?post=1179141"},{"taxonomy":"msr-publication-type","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-type?post=1179141"},{"taxonomy":"msr-publisher","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publisher?post=1179141"},{"taxonomy":"msr-publication-cta","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-publication-cta?post=1179141"},{"taxonomy":"msr-focus-area","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-focus-area?post=1179141"},{"taxonomy":"msr-locale","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-locale?post=1179141"},{"taxonomy":"msr-post-option","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-post-option?post=1179141"},{"taxonomy":"msr-field-of-study","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-field-of-study?post=1179141"},{"taxonomy":"msr-conference","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-conference?post=1179141"},{"taxonomy":"msr-journal","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-journal?post=1179141"},{"taxonomy":"msr-impact-theme","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-impact-theme?post=1179141"},{"taxonomy":"msr-pillar","embeddable":true,"href":"https:\/\/research.codeghost.online\/en-us\/research\/wp-json\/wp\/v2\/msr-pillar?post=1179141"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}