{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,7]],"date-time":"2026-10-07T18:44:57Z","timestamp":1791398697802,"version":"4.3.3"},"reference-count":33,"publisher":"Wiley","issue":"8","license":[{"start":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T00:00:00Z","timestamp":1778630400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T00:00:00Z","timestamp":1778630400000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["anatomypubs.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Anatomical Sciences Ed"],"published-print":{"date-parts":[[2026,8]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    Multimodal large language models (LLMs) are now deeply integrated into medical education and widely used by medical students, yet it remains unclear whether current models possess the accuracy and reliability needed to support image\u2010based learning. We evaluated four state\u2010of\u2010the\u2010art multimodal LLMs (ChatGPT\u20105.1, Gemini\u20102.5, Grok\u20104, Claude Sonnet\u20104.5) on 208 image\u2010based examination questions from a Doctor of Medicine program, spanning anatomical pathology (histopathology; 47.6%), radiology (31.7%), and surgical anatomy (20.7%). To isolate visual reasoning, all items were presented in image\u2010only form with contextual information removed. Items covered seven organ systems, included both constructed\u2010response and selected\u2010response formats, and were categorized as recognition\u2010only or recognition\u2010plus\u2010reasoning. ChatGPT\u20105.1 achieved the highest accuracy (75.5%; 95% CI [69.2\u201380.8]), followed by Gemini\u20102.5 (59.6%; 95% CI [52.8\u201366.1]), Claude Sonnet\u20104.5 (41.8%; 95% CI [35.3\u201348.6]), and Grok\u20104 (34.6%; 95% CI [28.5\u201341.3]). Overall model performance differed significantly (\n                    <jats:italic>p<\/jats:italic>\n                    \u2009&lt;\u20090.001; Cram\u00e9r's\n                    <jats:italic>V<\/jats:italic>\n                    \u2009=\u20090.45). Pairwise McNemar comparisons showed significant differences between ChatGPT\u20105.1 and all other models (all\n                    <jats:italic>p<\/jats:italic>\n                    \u2009&lt;\u20090.001; effect sizes 15.9\u201340.9 percentage points). Subgroup analyses demonstrated a consistent hierarchy (ChatGPT &gt; Gemini &gt; Claude \u2248 Grok) across different categories. Accuracy was uniformly higher for recognition\u2010only and selected\u2010response items. Even the best\u2010performing model, ChatGPT\u20105.1, answered approximately one in four questions incorrectly. These findings suggest that current multimodal LLMs cannot yet replace expert teaching in image\u2010based learning. Their use in medical education should therefore remain supervised and critically appraised, serving as adjuncts rather than authoritative sources.\n                  <\/jats:p>","DOI":"10.1002\/ase.70256","type":"journal-article","created":{"date-parts":[[2026,5,13]],"date-time":"2026-05-13T09:09:21Z","timestamp":1778663361000},"page":"1288-1299","update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["Performance of multimodal large language models on image\u2010based surgical anatomy, anatomical pathology, and radiology questions"],"prefix":"10.1002","volume":"19","author":[{"given":"Ming","family":"Lu","sequence":"first","affiliation":[{"name":"Medical Education, School of Medicine and Dentistry Griffith University  Southport Queensland Australia"},{"name":"Gold Coast University Hospital  Southport Queensland Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Josiah","family":"Cheng","sequence":"additional","affiliation":[{"name":"Medical Education, School of Medicine and Dentistry Griffith University  Southport Queensland Australia"},{"name":"Logan Hospital  Meadowbrook Queensland Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Vinod","family":"Gopalan","sequence":"additional","affiliation":[{"name":"Medical Education, School of Medicine and Dentistry Griffith University  Southport Queensland Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,5,13]]},"reference":[{"key":"e_1_2_13_2_1","doi-asserted-by":"publisher","DOI":"10.1186\/s12909-025-07438-7"},{"key":"e_1_2_13_3_1","doi-asserted-by":"crossref","DOI":"10.1177\/23821205251391969","article-title":"Perceptions and use of generative artificial intelligence in medical students: a multicenter survey","volume":"12","author":"Tran C","year":"2025","journal-title":"J Med Educ Curric Dev [Internet]"},{"key":"e_1_2_13_4_1","doi-asserted-by":"publisher","DOI":"10.2196\/77552"},{"key":"e_1_2_13_5_1","doi-asserted-by":"publisher","DOI":"10.2196\/60807"},{"key":"e_1_2_13_6_1","doi-asserted-by":"publisher","DOI":"10.2196\/50965"},{"key":"e_1_2_13_7_1","doi-asserted-by":"publisher","DOI":"10.2196\/63430"},{"key":"e_1_2_13_8_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-98949-2"},{"issue":"6","key":"e_1_2_13_9_1","doi-asserted-by":"crossref","DOI":"10.1371\/journal.pone.0324841","article-title":"Advancing medical AI: GPT\u20104 and GPT\u20104o surpass GPT\u20103.5 in Taiwanese medical licensing exams","volume":"20","author":"Wu YC","year":"2025","journal-title":"PLoS One [Internet]"},{"key":"e_1_2_13_10_1","doi-asserted-by":"publisher","DOI":"10.1016\/S0140-6736(10)61854-5"},{"key":"e_1_2_13_11_1","volume-title":"Basic medical education WFME global standards for quality improvement: The 2020 Revision [Internet]","year":"2020"},{"key":"e_1_2_13_12_1","doi-asserted-by":"publisher","DOI":"10.1093\/nsr\/nwae403"},{"issue":"9","key":"e_1_2_13_13_1","article-title":"A comprehensive review of multimodal large language models: performance and challenges across different tasks","volume":"18","author":"Wang J","year":"2024","journal-title":"arXivorg [Internet]"},{"key":"e_1_2_13_14_1","doi-asserted-by":"publisher","DOI":"10.2196\/65146"},{"key":"e_1_2_13_15_1","doi-asserted-by":"publisher","DOI":"10.1186\/s12909-025-07752-0"},{"key":"e_1_2_13_16_1","article-title":"Evaluating the performance of ChatGPT\u20104o vision capabilities on image\u2010based USMLE step 1, step 2, and step 3 examination questions","author":"Gajjar AA","year":"2024","journal-title":"bioRxiv (Cold Spring Harbor Laboratory)"},{"key":"e_1_2_13_17_1","doi-asserted-by":"publisher","DOI":"10.4274\/dir.2025.253460"},{"key":"e_1_2_13_18_1","doi-asserted-by":"publisher","DOI":"10.1002\/imo2.7"},{"key":"e_1_2_13_19_1","doi-asserted-by":"publisher","DOI":"10.5116\/ijme.625f.bfb1"},{"key":"e_1_2_13_20_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.yjoc.2025.100113"},{"key":"e_1_2_13_21_1","doi-asserted-by":"publisher","DOI":"10.1111\/j.1468-2958.2004.tb00738.x"},{"key":"e_1_2_13_22_1","doi-asserted-by":"publisher","DOI":"10.2466\/pms.104.4.1236-1242"},{"key":"e_1_2_13_23_1","volume-title":"The SAGE encyclopedia of educational research, measurement, and evaluation","author":"McHugh ML","year":"2018"},{"key":"e_1_2_13_24_1","doi-asserted-by":"publisher","DOI":"10.1080\/10543406.2013.789888"},{"key":"e_1_2_13_25_1","doi-asserted-by":"publisher","DOI":"10.1002\/ase.70074"},{"issue":"3","key":"e_1_2_13_26_1","doi-asserted-by":"crossref","first-page":"551","DOI":"10.32708\/uutfd.1568479","article-title":"Comparative performance evaluation of multimodal large language models, radiologist, and anatomist in visual neuroanatomy questions","volume":"50","author":"G\u00fcne\u015f YC","year":"2025","journal-title":"Uluda\u011f Univ Tip Fak Derg"},{"issue":"1","key":"e_1_2_13_27_1","article-title":"Large language models with vision on diagnostic radiology board exam style questions","volume":"313","author":"Sun SH","year":"2024","journal-title":"Acad Radiol"},{"key":"e_1_2_13_28_1","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.240273"},{"key":"e_1_2_13_29_1","article-title":"PathMMU: a massive multimodal expert\u2010level benchmark for understanding and reasoning in pathology","author":"Sun Y","year":"2024","journal-title":"arXiv (Cornell University)"},{"key":"e_1_2_13_30_1","doi-asserted-by":"publisher","DOI":"10.3390\/bioengineering10030380"},{"key":"e_1_2_13_31_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-22940-0"},{"key":"e_1_2_13_32_1","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-025-97327-2"},{"key":"e_1_2_13_33_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2025.110614"},{"key":"e_1_2_13_34_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.patter.2025.101260"}],"container-title":["Anatomical Sciences Education"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/anatomypubs.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/ase.70256","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/anatomypubs.onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/ase.70256","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/anatomypubs.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/ase.70256","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,5]],"date-time":"2026-08-05T17:39:34Z","timestamp":1785951574000},"score":1,"resource":{"primary":{"URL":"https:\/\/anatomypubs.onlinelibrary.wiley.com\/doi\/10.1002\/ase.70256"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,13]]},"references-count":33,"journal-issue":{"issue":"8","published-print":{"date-parts":[[2026,8]]}},"alternative-id":["10.1002\/ase.70256"],"URL":"https:\/\/doi.org\/10.1002\/ase.70256","archive":["Portico"],"relation":{},"ISSN":["1935-9772","1935-9780"],"issn-type":[{"value":"1935-9772","type":"print"},{"value":"1935-9780","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,5,13]]},"assertion":[{"value":"2025-12-02","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-04-24","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-05-13","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}