{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,4]],"date-time":"2026-08-04T04:04:43Z","timestamp":1785816283962,"version":"3.56.0"},"reference-count":16,"publisher":"American Medical Association (AMA)","issue":"7","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JAMA Dermatol"],"published-print":{"date-parts":[[2026,7,1]]},"abstract":"<jats:sec id=\"ab-doi260018-4\">\n                    <jats:title>Importance<\/jats:title>\n                    <jats:p>Artificial intelligence (AI) systems for skin cancer detection perform well in controlled settings but frequently underperform in everyday clinical practice, raising critical questions about their readiness for deployment.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-5\">\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>To compare the diagnostic accuracy of AI algorithms vs human evaluators across varying expertise levels for skin lesion diagnosis, including rare and atypical cases, in a realistic clinical context.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-6\">\n                    <jats:title>Design, Setting, and Participants<\/jats:title>\n                    <jats:p>This multi-institutional diagnostic study compared diagnostic performance among AI models and physician readers with varying dermatological expertise, ranging from less than 1 year to more than 10 years of experience. A dataset of dermatological images representing everyday clinical scenarios was used and contained 1117 cases, including clinical and dermoscopic images with associated metadata. Study inclusion spanned from March 16, 2023, to August 1, 2025.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-7\">\n                    <jats:title>Exposures<\/jats:title>\n                    <jats:p>Three AI algorithms: a first-generation convolutional neural network (CNN) and 2 foundation models (PanDerm unimodal and multimodal). Human readers evaluated 100 stratified, random cases drawn from the same dataset.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-8\">\n                    <jats:title>Main Outcomes and Measures<\/jats:title>\n                    <jats:p>The primary outcome was reader-level multiclass diagnostic accuracy for skin lesion classification. Secondary outcomes were binary benign vs malignant sensitivity, specificity, and balanced accuracy. Performance was compared between AI algorithms and human readers stratified by experience level.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-9\">\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      A total of 652 physicians (median [IQR] age, 33 [29-37] years; 559 [85.7%] female) contributed to 1092 test iterations. All human readers outperformed the CNN (mean [SD] accuracy, 65.9% [10.5%] vs 56.7% [3.9%]; difference, 9.2 percentage points [pp]; 95% CI, \u22129.8 to 8.5 pp;\n                      <jats:italic toggle=\"yes\">P<\/jats:italic>\n                      \u2009&amp;amp;lt;\u2009.001). Unimodal accuracy exceeded readers with less than 3 years of experience (mean [SD] accuracy, 72.2% [3.5%] vs 68.2% [7.6%]; difference, 4.0 pp; 95% CI, 3.2-4.9 pp;\n                      <jats:italic toggle=\"yes\">P<\/jats:italic>\n                      \u2009&amp;amp;lt;\u2009.001). With a mean (SD) accuracy of 74.2% (5.7%), experts with more than 10 years of experience achieved the highest multiclass diagnostic accuracy, outperforming all AI models on this primary end point, which included 56.7% (3.9%) for CNN, 72.2% (3.5%) for the unimodal model, and 66.3% (3.8%) for the multimodal model.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec id=\"ab-doi260018-10\">\n                    <jats:title>Conclusions and Relevance<\/jats:title>\n                    <jats:p>In this diagnostic study, a modern foundation model surpassed readers with less than 3 years of experience on accuracy of skin lesion diagnosis and matched those with 3 to 10 years of experience but remained inferior to experts with more than 10 years of experience, highlighting both the promise and current limitations of AI in dermatologic diagnosis.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1001\/jamadermatol.2026.1492","type":"journal-article","created":{"date-parts":[[2026,6,3]],"date-time":"2026-06-03T15:30:36Z","timestamp":1780500636000},"page":"701","source":"Crossref","is-referenced-by-count":1,"title":["Limits of Artificial Intelligence Models for Skin Cancer Diagnosis in Realistic Settings"],"prefix":"10.1001","volume":"162","author":[{"given":"Julien","family":"Anriot","sequence":"first","affiliation":[{"name":"Claude Bernard University Lyon 1, Lyon, France"},{"name":"Centre L\u00e9on Berard, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Siyuan","family":"Yan","sequence":"additional","affiliation":[{"name":"AIM for Health Lab, Faculty of Information Technology, Monash University, Melbourne, Victoria, Australia"},{"name":"Faculty of Engineering, Monash University, Melbourne, Victoria, Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Clio","family":"Coste","sequence":"additional","affiliation":[{"name":"Centre L\u00e9on Berard, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Philipp","family":"Tschandl","sequence":"additional","affiliation":[{"name":"Department of Dermatology, Medical University of Vienna, Vienna, Austria"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Loic","family":"Verlingue","sequence":"additional","affiliation":[{"name":"Centre L\u00e9on Berard, Lyon, France"},{"name":"Cancer Research Center of Lyon, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Camille","family":"Andremasse","sequence":"additional","affiliation":[{"name":"Claude Bernard University Lyon 1, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mona","family":"Amini-Adle","sequence":"additional","affiliation":[{"name":"Centre L\u00e9on Berard, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jean Luc","family":"Perrot","sequence":"additional","affiliation":[{"name":"Jean Moulin University, St Etienne, France"},{"name":"Laboratoire H Curien UMR CNRS 5516, St Etienne, France"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zongyuan","family":"Ge","sequence":"additional","affiliation":[{"name":"AIM for Health Lab, Faculty of Information Technology, Monash University, Melbourne, Victoria, Australia"},{"name":"Airdoc-Monash Research, Monash University, Melbourne, Victoria, Australia"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Harald","family":"Kittler","sequence":"additional","affiliation":[{"name":"Department of Dermatology, Medical University of Vienna, Vienna, Austria"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Luc","family":"Thomas","sequence":"additional","affiliation":[{"name":"Claude Bernard University Lyon 1, Lyon, France"},{"name":"Cancer Research Center of Lyon, Lyon, France"},{"name":"Dermatology Department, H\u00f4pital Lyon Sud, Hospices Civils de Lyon, Lyon, France"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"10","reference":[{"issue":"8","key":"doi260018r1","doi-asserted-by":"publisher","first-page":"1836","DOI":"10.1093\/annonc\/mdy166","article-title":"Man against machine: diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists.","volume":"29","author":"Haenssle","year":"2018","journal-title":"Ann Oncol"},{"key":"doi260018r2","doi-asserted-by":"publisher","first-page":"202","DOI":"10.1016\/j.ejca.2021.06.049","article-title":"Skin cancer classification via convolutional neural networks: systematic review of studies involving human experts.","volume":"156","author":"Haggenm\u00fcller","year":"2021","journal-title":"Eur J Cancer"},{"issue":"11","key":"doi260018r3","doi-asserted-by":"publisher","first-page":"265","DOI":"10.3390\/jimaging10110265","article-title":"Skin cancer image classification using artificial intelligence strategies: a systematic review.","volume":"10","author":"Vardasca","year":"2024","journal-title":"J Imaging"},{"issue":"3","key":"doi260018r4","doi-asserted-by":"publisher","first-page":"280","DOI":"10.1016\/j.clindermatol.2023.12.022","article-title":"Artificial intelligence in the detection of skin cancer: state of the art.","volume":"42","author":"Strzelecki","year":"2024","journal-title":"Clin Dermatol"},{"key":"doi260018r5","doi-asserted-by":"publisher","DOI":"10.1016\/j.ejca.2023.112954","article-title":"Diagnostic performance of augmented intelligence with 2D and 3D total body photography and convolutional neural networks in a high-risk population for melanoma under real-world conditions: a new era of skin cancer screening?","volume":"190","author":"Cerminara","year":"2023","journal-title":"Eur J Cancer"},{"issue":"7","key":"doi260018r6","doi-asserted-by":"publisher","first-page":"e607","DOI":"10.1111\/jdv.20573","article-title":"Creation and evaluation of an online dermoscopy test for self-evaluation of competences in diagnosis of skin cancer.","volume":"39","author":"Andremasse","year":"2025","journal-title":"J Eur Acad Dermatol Venereol"},{"key":"doi260018r7","doi-asserted-by":"publisher","DOI":"10.1038\/sdata.2018.161","article-title":"The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions.","volume":"5","author":"Tschandl","year":"2018","journal-title":"Sci Data"},{"issue":"8","key":"doi260018r8","doi-asserted-by":"publisher","first-page":"2691","DOI":"10.1038\/s41591-025-03747-y","article-title":"A multimodal vision foundation model for clinical dermatology.","volume":"31","author":"Yan","year":"2025","journal-title":"Nat Med"},{"issue":"1","key":"doi260018r9","doi-asserted-by":"publisher","first-page":"90","DOI":"10.1001\/jamadermatol.2021.4915","article-title":"Checklist for Evaluation of Image-Based Artificial Intelligence Reports in Dermatology: CLEAR Derm consensus guidelines from the International Skin Imaging Collaboration Artificial Intelligence Working Group.","volume":"158","author":"Daneshjou","year":"2022","journal-title":"JAMA Dermatol"},{"issue":"7","key":"doi260018r10","doi-asserted-by":"publisher","first-page":"938","DOI":"10.1016\/S1470-2045(19)30333-X","article-title":"Comparison of the accuracy of human readers versus machine-learning algorithms for pigmented skin lesion classification: an open, web-based, international, diagnostic study.","volume":"20","author":"Tschandl","year":"2019","journal-title":"Lancet Oncol"},{"issue":"3","key":"doi260018r11","doi-asserted-by":"publisher","first-page":"180","DOI":"10.1097\/CMR.0b013e32832a1e41","article-title":"Computer versus human diagnosis of melanoma: evaluation of the feasibility of an automated diagnostic system in a prospective clinical trial.","volume":"19","author":"Dreiseitl","year":"2009","journal-title":"Melanoma Res"},{"issue":"4","key":"doi260018r12","doi-asserted-by":"publisher","first-page":"813","DOI":"10.1016\/j.jaad.2016.03.027","article-title":"The study of nevi in children: principles learned and implications for melanoma diagnosis.","volume":"75","author":"Scope","year":"2016","journal-title":"J Am Acad Dermatol"},{"issue":"1","key":"doi260018r13","doi-asserted-by":"publisher","first-page":"34","DOI":"10.1038\/s41597-021-00815-z","article-title":"A patient-centric dataset of images and metadata for identifying melanomas using clinical context.","volume":"8","author":"Rotemberg","year":"2021","journal-title":"Sci Data"},{"issue":"3","key":"doi260018r14","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0193321","article-title":"Acral melanoma detection using a convolutional neural network for dermoscopy images.","volume":"13","author":"Yu","year":"2018","journal-title":"PLoS One"},{"issue":"12","key":"doi260018r15","doi-asserted-by":"publisher","first-page":"2323","DOI":"10.1111\/jdv.20328","article-title":"Impact of dermoscopy training associated with artificial intelligence on general practitioner residents.","volume":"38","author":"Dorado Cortez","year":"2024","journal-title":"J Eur Acad Dermatol Venereol"},{"issue":"10","key":"doi260018r16","doi-asserted-by":"publisher","first-page":"896","DOI":"10.1016\/S2468-1253(25)00133-5","article-title":"Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study.","volume":"10","author":"Budzyn","year":"2025","journal-title":"Lancet Gastroenterol Hepatol"}],"container-title":["JAMA Dermatology"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/jamanetwork.com\/journals\/jamadermatology\/articlepdf\/2849416\/jamadermatology_anriot_2026_oi_260018_1783528897.34522.pdf","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,15]],"date-time":"2026-07-15T15:00:39Z","timestamp":1784127639000},"score":1,"resource":{"primary":{"URL":"https:\/\/jamanetwork.com\/journals\/jamadermatology\/fullarticle\/2849416"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,1]]},"references-count":16,"journal-issue":{"issue":"7","published-print":{"date-parts":[[2026,7,1]]}},"URL":"https:\/\/doi.org\/10.1001\/jamadermatol.2026.1492","relation":{},"ISSN":["2168-6068"],"issn-type":[{"value":"2168-6068","type":"print"}],"subject":[],"published":{"date-parts":[[2026,7,1]]}}}