{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T20:54:29Z","timestamp":1790888069215,"version":"4.1.0"},"reference-count":28,"publisher":"American Society of Clinical Oncology (ASCO)","issue":"4","content-domain":{"domain":["ascopubs.org"],"crossmark-restriction":true},"short-container-title":["JCO Clin Cancer Inform"],"published-print":{"date-parts":[[2026,7]]},"abstract":"<jats:sec>\n                    <jats:title>BACKGROUND AND SCOPE<\/jats:title>\n                    <jats:p>Cause of death is an important variable in cancer registries that supports downstream analyses, including cause-specific survival analysis. Unfortunately, in current practice, it is only available on government-issued death certificates, which are often received as scanned images. Current computer vision-based optical character recognition (OCR) methods struggle with handwriting, background noise, and blurriness, resulting in poor performance. A robust method for the cause of death ascertainment is urgently needed.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>SOLUTION<\/jats:title>\n                    <jats:p>Latest vision language models (VLMs), such as Qwen3-VL and GPT-4.1, have shown promising OCR performance on poor-quality scanned documents in the general domain. We proposed two prompting methods: (1) the two-step approach, which prompts VLMs for a full-text OCR followed by a text-based information extraction and (2) end-to-end approach, which prompts VLMs with a JavaScript Object Notation schema for image-based information extraction.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>EVALUATION<\/jats:title>\n                    <jats:p>\n                      We evaluated VLMs for the cause of death extraction from scanned government-issued death certificates in a major cancer center. Of 22,733 death records, 500 were manually annotated as ground truth. For the immediate cause of death,\n                      <jats:italic toggle=\"yes\">Qwen3-VL-30B-A3B-Thinking<\/jats:italic>\n                      with the end-to-end approach achieved the best accuracy of 0.8625, significantly outperforming the Tesseract baseline (0.2396). For the underlying causes of death, the two-step method achieved the best F1 score of 0.7637, outperforming the baseline (0.2451).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>RELEVANCE<\/jats:title>\n                    <jats:p>The growing emphasis on precision medicine underscores the importance of accurate cause of death ascertainment for survival analyses. Our method demonstrated significant improvements over the current OCR approaches and provided a reliable alternative to reliance on the National Death Index.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>HOW TO USE<\/jats:title>\n                    <jats:p>Our VLM OCR computational framework is publicly available on GitHub. The open-weight VLMs are available for download on the Hugging Face repository.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1200\/cci-26-00037","type":"journal-article","created":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T20:00:12Z","timestamp":1790884812000},"update-policy":"https:\/\/doi.org\/10.1200\/crossmark","source":"Crossref","is-referenced-by-count":0,"title":["Enhancing Cancer Outcomes Research Through Vision Language Models for Cause of Death Ascertainment"],"prefix":"10.1200","volume":"10","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8137-0171","authenticated-orcid":true,"given":"Enshuo","family":"Hsu","sequence":"first","affiliation":[{"name":"Enterprise Development and Integration, University of Texas MD Anderson Cancer Center, Houston, TX"},{"name":"McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Giordana","family":"De Las Pozas","sequence":"additional","affiliation":[{"name":"Department of Cancer Registry, University of Texas MD Anderson Cancer Center, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Trey","family":"Kell","sequence":"additional","affiliation":[{"name":"Enterprise Development and Integration, University of Texas MD Anderson Cancer Center, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6525-5213","authenticated-orcid":true,"given":"Kirk","family":"Roberts","sequence":"additional","affiliation":[{"name":"McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7252-9702","authenticated-orcid":true,"given":"Shida","family":"Jin","sequence":"additional","affiliation":[{"name":"Enterprise Development and Integration, University of Texas MD Anderson Cancer Center, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4727-6269","authenticated-orcid":true,"given":"Kelly W.","family":"Merriman","sequence":"additional","affiliation":[{"name":"McWilliams School of Biomedical Informatics, University of Texas Health Science Center at Houston, Houston, TX"},{"name":"Department of Cancer Registry, University of Texas MD Anderson Cancer Center, Houston, TX"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"233","reference":[{"key":"e_1_3_4_2_2","doi-asserted-by":"publisher","DOI":"10.1245\/s10434-024-15393-8"},{"key":"e_1_3_4_3_2","volume-title":"Data Standards and Data Dictionary","author":"North American Association of Central Cancer Registries","year":"2021","unstructured":"North American Association of Central Cancer Registries: Data Standards and Data Dictionary, 2021"},{"key":"e_1_3_4_4_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.gie.2020.08.038"},{"key":"e_1_3_4_5_2","volume-title":"Tesseract-Ocr\/Tesseract","year":"2025","unstructured":"Tesseract-Ocr\/Tesseract. 2025"},{"key":"e_1_3_4_6_2","doi-asserted-by":"publisher","DOI":"10.18100\/ijamec.270374"},{"key":"e_1_3_4_7_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpb.2019.01.012"},{"key":"e_1_3_4_8_2","doi-asserted-by":"publisher","DOI":"10.1093\/jamiaopen\/ooac045"},{"key":"e_1_3_4_9_2","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijmedinf.2020.104302"},{"key":"e_1_3_4_10_2","doi-asserted-by":"publisher","unstructured":"Buckley T Diao JA Rajpurkar P et al: Multimodal foundation models exploit text to make medical image predictions. ArXiv 10.48550\/arXiv.2311.05591","DOI":"10.48550\/arXiv.2311.05591"},{"key":"e_1_3_4_11_2","doi-asserted-by":"publisher","DOI":"10.3389\/frai.2024.1430984"},{"key":"e_1_3_4_12_2","doi-asserted-by":"publisher","unstructured":"Lin T Zhang W Li S et al: Healthgpt: A medical large vision-language model for unifying comprehension and generation via heterogeneous knowledge adaptation ArXiv 10.48550\/arXiv.2502.09838","DOI":"10.48550\/arXiv.2502.09838"},{"key":"e_1_3_4_13_2","doi-asserted-by":"publisher","DOI":"10.1038\/s41591-024-03185-2"},{"key":"e_1_3_4_14_2","doi-asserted-by":"publisher","DOI":"10.2196\/32690"},{"key":"e_1_3_4_15_2","doi-asserted-by":"publisher","unstructured":"Sch\u00e4fer H Schmidt CS Wutzkowsky J et al: A multimodal pipeline for clinical data extraction: Applying vision-language models to scans of transfusion reaction reports. ArXiv 10.48550\/arXiv.2504.20220","DOI":"10.48550\/arXiv.2504.20220"},{"key":"e_1_3_4_16_2","doi-asserted-by":"publisher","unstructured":"Yang A Li A Yang B et al: Qwen3 Technical Report ArXiv 10.48550\/arXiv.2505.09388","DOI":"10.48550\/arXiv.2505.09388"},{"key":"e_1_3_4_17_2","volume-title":"Qwen2.5-VL Technical Report","author":"Bai S","year":"2025","unstructured":"Bai S, Chen K, Liu X, et al: Qwen2.5-VL Technical Report, 2025"},{"key":"e_1_3_4_18_2","doi-asserted-by":"publisher","unstructured":"Achiam J Adler S Agarwal S et al: Gpt-4 Technical Report ArXiv 10.48550\/arXiv.2303.08774","DOI":"10.48550\/arXiv.2303.08774"},{"key":"e_1_3_4_19_2","unstructured":"Introducing GPT-4.1 in the API. https:\/\/openai.com\/index\/gpt-4-1\/"},{"key":"e_1_3_4_20_2","doi-asserted-by":"publisher","DOI":"10.1093\/jamiaopen\/ooaf012"},{"key":"e_1_3_4_21_2","unstructured":"openai\/gpt-oss: gpt-oss-120b and gpt-oss-20b are two open-weight language models by OpenAI. https:\/\/github.com\/openai\/gpt-oss"},{"key":"e_1_3_4_22_2","doi-asserted-by":"publisher","DOI":"10.1145\/3600006.3613165"},{"key":"e_1_3_4_23_2","volume-title":"daviden1013\/vlm4ocr","author":"Hsu D","year":"2025","unstructured":"Hsu D: daviden1013\/vlm4ocr, 2025"},{"key":"e_1_3_4_24_2","doi-asserted-by":"publisher","DOI":"10.1080\/02664769723918"},{"key":"e_1_3_4_25_2","unstructured":"CDC: National death index 2026. https:\/\/www.cdc.gov\/nchs\/ndi\/index.html"},{"key":"e_1_3_4_26_2","unstructured":"OpenRouter: OpenRouter. https:\/\/openrouter.ai"},{"key":"e_1_3_4_27_2","unstructured":"VLM4OCR Documentation: daviden1013\/vlm4ocr 2026. https:\/\/daviden1013.github.io\/vlm4ocr\/https:\/\/github.com\/daviden1013\/vlm4ocr"},{"key":"e_1_3_4_28_2","unstructured":"VLM4OCR: https:\/\/daviden1013.github.io\/vlm4ocr\/"},{"key":"e_1_3_4_29_2","unstructured":"Bai S Cai Y Chen R et al: Qwen3-vl technical report. arXiv:2511.21631"}],"container-title":["JCO Clinical Cancer Informatics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/ascopubs.org\/doi\/pdfdirect\/10.1200\/CCI-26-00037","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T20:00:14Z","timestamp":1790884814000},"score":1,"resource":{"primary":{"URL":"https:\/\/ascopubs.org\/doi\/10.1200\/CCI-26-00037"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7]]},"references-count":28,"journal-issue":{"issue":"4","published-print":{"date-parts":[[2026,7]]}},"alternative-id":["10.1200\/CCI-26-00037"],"URL":"https:\/\/doi.org\/10.1200\/cci-26-00037","relation":{},"ISSN":["2473-4276"],"issn-type":[{"value":"2473-4276","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7]]},"assertion":[{"value":"2026-02-09","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-07-07","order":1,"name":"revised","label":"Revised","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-07-31","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-10-01","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e2600037"}}