{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,14]],"date-time":"2026-10-14T02:45:02Z","timestamp":1791945902330,"version":"4.3.4"},"reference-count":31,"publisher":"Springer Science and Business Media LLC","license":[{"start":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T00:00:00Z","timestamp":1791849600000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T00:00:00Z","timestamp":1791849600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["J Digit Imaging. Inform. med."],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Frontier large language models (LLMs) and diffusion models can generate realistic chest radiographs, raising concerns for dataset integrity when provenance metadata are absent. We compared authenticity judgments by two radiologists, two nonradiologist physicians, and three frontier LLMs (GPT-5.5, Gemini 3.5 Flash, Claude Opus 4.8) on 160 posteroanterior radiographs: 40 real NIH ChestX-ray14 images and 120 synthetic images generated from text by three pipelines. Each evaluator judged every image once; the LLM evaluation was repeated twice 3\u00a0months later. Balanced accuracy was the primary metric, given 75% synthetic prevalence. The two best evaluators reached nearly identical balanced accuracy from opposite operating points: Gemini 3.5 Flash had the highest point estimate (77.9%; specificity 100%, sensitivity 55.8%) and RAD1 was a close second (75.8%; specificity 55.0%, sensitivity 96.7%), a paired difference of\u2009\u2212\u20092.1 percentage points (95% CI,\u2009\u2212\u20095.8 to 1.2). On repetition, LLM judgments were not fully reproducible and this ordering did not hold. Nonradiologists performed near chance. All four human readers detected diffusion-generated images at 82\u2013100%, whereas the radiologists misclassified 38\u201345% of real images and the conservative LLMs reached perfect specificity yet detected only 32\u201372% of LLM-generated images. With laterality markers and borders masked, specificity was 62.5% (Gemini) and 85.0% (GPT-5.5) against 90\u2013100% in same-day unmasked runs. Interevaluator agreement was low throughout (Fleiss\u2019 kappa among LLMs 0.213). A frequency-domain baseline reached 98.3% balanced accuracy within this dataset but 12% sensitivity on a held-out generator. Human readers and LLMs failed on different images; visual authenticity assessment alone may be insufficient for provenance verification.<\/jats:p>","DOI":"10.1007\/s10278-026-02339-7","type":"journal-article","created":{"date-parts":[[2026,10,13]],"date-time":"2026-10-13T20:46:34Z","timestamp":1791924394000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Complementary Error Profiles of Human Readers and Large Language Models in Detecting AI-Generated Chest Radiographs: A Comparative Reader Study"],"prefix":"10.1007","author":[{"given":"Suji","family":"Lee","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jei Youn","family":"Park","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yongsik","family":"Sim","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2469-839X","authenticated-orcid":false,"given":"Youho","family":"Myong","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,10,13]]},"reference":[{"key":"2339_CR1","doi-asserted-by":"publisher","DOI":"10.1016\/j.media.2023.102846","volume":"88","author":"A Kazerouni","year":"2023","unstructured":"Kazerouni A, Aghdam EK, Heidari M, Azad R, Fayyaz M, Hacihaliloglu I, et al. Diffusion models in medical imaging: A comprehensive survey. Medical image analysis. 2023;88:102846.","journal-title":"Medical image analysis."},{"key":"2339_CR2","first-page":"6840","volume":"33","author":"J Ho","year":"2020","unstructured":"Ho J, Jain A, Abbeel P. Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems. 2020;33:6840-51.","journal-title":"Advances in Neural Information Processing Systems."},{"key":"2339_CR3","doi-asserted-by":"crossref","unstructured":"Rombach R, Blattmann A, Lorenz D, Esser P, Ommer B, editors. High-resolution image synthesis with latent diffusion models. Proceedings of the IEEE\/CVF conference on computer vision and pattern recognition; 2022.","DOI":"10.1109\/CVPR52688.2022.01042"},{"key":"2339_CR4","doi-asserted-by":"publisher","DOI":"10.1016\/j.neuroimage.2024.120663","volume":"296","author":"D Yoon","year":"2024","unstructured":"Yoon D, Myong Y, Kim YG, Sim Y, Cho M, Oh B-M, et al. Latent diffusion model-based MRI superresolution enhances mild cognitive impairment prognostication and Alzheimer\u2019s disease classification. Neuroimage. 2024;296:120663.","journal-title":"Neuroimage."},{"key":"2339_CR5","unstructured":"Chambon P, Bluethgen C, Delbrouck J-B, Van der Sluijs R, Po\u0142acin M, Chaves JMZ, et al. Roentgen: vision-language foundation model for chest x-ray generation. arXiv preprint arXiv:2211.12737. 2022."},{"key":"2339_CR6","doi-asserted-by":"crossref","unstructured":"Khan W, Leem S, See KB, Wong JK, Zhang S, Fang R. A comprehensive survey of foundation models in medicine. IEEE Reviews in Biomedical Engineering. 2025.","DOI":"10.1109\/RBME.2025.3531360"},{"issue":"4","key":"2339_CR7","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pone.0279349","volume":"18","author":"Y Myong","year":"2023","unstructured":"Myong Y, Yoon D, Kim BS, Kim YG, Sim Y, Lee S, et al. Evaluating diagnostic content of AI-generated chest radiography: A multi-center visual Turing test. PLoS One. 2023;18(4):e0279349.","journal-title":"PLoS One."},{"key":"2339_CR8","doi-asserted-by":"crossref","unstructured":"Jang M, Bae H-j, Kim M, Park SY, Son A-Y, Choi SJ, et al. Image Turing test and its applications on synthetic chest radiographs by using the progressive growing generative adversarial network. Scientific reports. 2023;13(1):2356.","DOI":"10.1038\/s41598-023-28175-1"},{"issue":"8022","key":"2339_CR9","doi-asserted-by":"publisher","first-page":"755","DOI":"10.1038\/s41586-024-07566-y","volume":"631","author":"I Shumailov","year":"2024","unstructured":"Shumailov I, Shumaylov Z, Zhao Y, Papernot N, Anderson R, Gal Y. AI models collapse when trained on recursively generated data. Nature. 2024;631(8022):755-9.","journal-title":"Nature."},{"key":"2339_CR10","doi-asserted-by":"crossref","unstructured":"He H, Xiang S, Zhang Y, Zhu Y, Zhang J, Lu Y, et al. AI-generated data contamination erodes pathological variability and diagnostic reliability. medRxiv. 2026:2026.01. 19.26344383.","DOI":"10.64898\/2026.01.19.26344383"},{"issue":"6","key":"2339_CR11","doi-asserted-by":"publisher","first-page":"695","DOI":"10.1007\/s10278-017-9961-x","volume":"30","author":"G Ulutas","year":"2017","unstructured":"Ulutas G, Ustubioglu A, Ustubioglu B, Nabiyev VV, Ulutas M. Medical image tamper detection based on passive image authentication. Journal of digital imaging. 2017;30(6):695-709.","journal-title":"Journal of digital imaging."},{"issue":"3","key":"2339_CR12","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.252094","volume":"318","author":"M Tordjman","year":"2026","unstructured":"Tordjman M, Yuce M, Ammar A, Huang M, Mihoubi Bouvier F, Lacroix M, et al. The rise of deepfake medical imaging: radiologists\u2019 diagnostic accuracy in detecting ChatGPT-generated radiographs. Radiology. 2026;318(3):e252094.","journal-title":"Radiology."},{"issue":"2","key":"2339_CR13","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.233270","volume":"311","author":"Y Zhou","year":"2024","unstructured":"Zhou Y, Ong H, Kennedy P, Wu CC, Kazam J, Hentel K, et al. Evaluating GPT-4V (GPT-4 with vision) on detection of radiologic findings on chest radiographs. Radiology. 2024;311(2):e233270.","journal-title":"Radiology."},{"issue":"4","key":"2339_CR14","doi-asserted-by":"publisher","first-page":"1959","DOI":"10.1007\/s00330-024-11035-5","volume":"35","author":"D Brin","year":"2025","unstructured":"Brin D, Sorin V, Barash Y, Konen E, Glicksberg BS, Nadkarni GN, et al. Assessing GPT-4 multimodal performance in radiological image analysis. European Radiology. 2025;35(4):1959-65.","journal-title":"European Radiology."},{"issue":"7","key":"2339_CR15","doi-asserted-by":"publisher","first-page":"4096","DOI":"10.1097\/JS9.0000000000001359","volume":"110","author":"L Zhu","year":"2024","unstructured":"Zhu L, Mou W, Lai Y, Chen J, Lin S, Xu L, et al. Step into the era of large multimodal models: a pilot study on ChatGPT-4V (ision)\u2019s ability to interpret radiological images. International Journal of Surgery. 2024;110(7):4096-102.","journal-title":"International Journal of Surgery."},{"issue":"10","key":"2339_CR16","doi-asserted-by":"publisher","first-page":"1168","DOI":"10.1007\/s11604-024-01606-3","volume":"42","author":"R Dehdab","year":"2024","unstructured":"Dehdab R, Brendlin A, Werner S, Almansour H, Gassenmaier S, Brendel JM, et al. Evaluating ChatGPT-4V in chest CT diagnostics: a critical image interpretation assessment. Japanese Journal of Radiology. 2024;42(10):1168-77.","journal-title":"Japanese Journal of Radiology."},{"issue":"1","key":"2339_CR17","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1175\/1520-0493(1950)078<0001:VOFEIT>2.0.CO;2","volume":"78","author":"GW Brier","year":"1950","unstructured":"Brier GW. Verification of forecasts expressed in terms of probability. Monthly weather review. 1950;78(1):1-3.","journal-title":"Monthly weather review."},{"key":"2339_CR18","doi-asserted-by":"publisher","first-page":"68772","DOI":"10.52202\/079017-2197","volume":"37","author":"A Panickssery","year":"2024","unstructured":"Panickssery A, Bowman SR, Feng S. Llm evaluators recognize and favor their own generations. Advances in Neural Information Processing Systems. 2024;37:68772-802.","journal-title":"Advances in Neural Information Processing Systems."},{"key":"2339_CR19","doi-asserted-by":"publisher","first-page":"46595","DOI":"10.52202\/075280-2020","volume":"36","author":"L Zheng","year":"2023","unstructured":"Zheng L, Chiang W-L, Sheng Y, Zhuang S, Wu Z, Zhuang Y, et al. Judging llm-as-a-judge with mt-bench and chatbot arena. Advances in neural information processing systems. 2023;36:46595-623.","journal-title":"Advances in neural information processing systems."},{"key":"2339_CR20","unstructured":"C2PA. Coalition for Content Provenance and Authenticity. 2024."},{"key":"2339_CR21","unstructured":"DeepMind G. SynthID: identifying AI-generated content 2024 [cited 2026 Jul 8]. Available from: https:\/\/deepmind.google\/technologies\/synthid\/."},{"key":"2339_CR22","doi-asserted-by":"crossref","unstructured":"Wen Y, Kirchenbauer J, Geiping J, Goldstein T. Tree-ring watermarks: Fingerprints for diffusion images that are invisible and robust. arXiv preprint arXiv:2305.20030. 2023.","DOI":"10.52202\/075280-2529"},{"key":"2339_CR23","doi-asserted-by":"crossref","unstructured":"Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM, editors. Chestx-ray8: Hospital-scale chest x-ray database and benchmarks on weakly-supervised classification and localization of common thorax diseases. Proceedings of the IEEE conference on computer vision and pattern recognition; 2017.","DOI":"10.1109\/CVPR.2017.369"},{"key":"2339_CR24","doi-asserted-by":"crossref","unstructured":"Brodersen KH, Ong CS, Stephan KE, Buhmann JM, editors. The balanced accuracy and its posterior distribution. 2010 20th international conference on pattern recognition; 2010: IEEE.","DOI":"10.1109\/ICPR.2010.764"},{"issue":"1","key":"2339_CR25","doi-asserted-by":"publisher","first-page":"32","DOI":"10.1002\/1097-0142(1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3","volume":"3","author":"WJ Youden","year":"1950","unstructured":"Youden WJ. Index for rating diagnostic tests. Cancer. 1950;3(1):32-5.","journal-title":"Cancer."},{"key":"2339_CR26","doi-asserted-by":"crossref","unstructured":"Matthews BW. Comparison of the predicted and observed secondary structure of T4 phage lysozyme. Biochimica et Biophysica Acta (BBA)-Protein Structure. 1975;405(2):442\u201351.","DOI":"10.1016\/0005-2795(75)90109-9"},{"issue":"1","key":"2339_CR27","doi-asserted-by":"publisher","first-page":"37","DOI":"10.1177\/001316446002000104","volume":"20","author":"J Cohen","year":"1960","unstructured":"Cohen J. A coefficient of agreement for nominal scales. Educational and psychological measurement. 1960;20(1):37-46.","journal-title":"Educational and psychological measurement."},{"issue":"5","key":"2339_CR28","doi-asserted-by":"publisher","first-page":"378","DOI":"10.1037\/h0031619","volume":"76","author":"JL Fleiss","year":"1971","unstructured":"Fleiss JL. Measuring nominal scale agreement among many raters. Psychological bulletin. 1971;76(5):378.","journal-title":"Psychological bulletin."},{"issue":"1","key":"2339_CR29","doi-asserted-by":"publisher","first-page":"29","DOI":"10.1348\/000711006X126600","volume":"61","author":"KL Gwet","year":"2008","unstructured":"Gwet KL. Computing inter\u2010rater reliability and its variance in the presence of high agreement. British Journal of Mathematical and Statistical Psychology. 2008;61(1):29-48.","journal-title":"British Journal of Mathematical and Statistical Psychology."},{"issue":"2","key":"2339_CR30","doi-asserted-by":"publisher","first-page":"153","DOI":"10.1007\/BF02295996","volume":"12","author":"Q McNemar","year":"1947","unstructured":"McNemar Q. Note on the sampling error of the difference between correlated proportions or percentages. Psychometrika. 1947;12(2):153-7.","journal-title":"Psychometrika."},{"key":"2339_CR31","unstructured":"Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine learning in Python. the Journal of machine Learning research. 2011;12:2825\u201330."}],"container-title":["Journal of Imaging Informatics in Medicine"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s10278-026-02339-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s10278-026-02339-7","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s10278-026-02339-7.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,14]],"date-time":"2026-10-14T02:01:57Z","timestamp":1791943317000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s10278-026-02339-7"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,13]]},"references-count":31,"alternative-id":["2339"],"URL":"https:\/\/doi.org\/10.1007\/s10278-026-02339-7","relation":{},"ISSN":["2948-2933"],"issn-type":[{"value":"2948-2933","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,13]]},"assertion":[{"value":"13 July 2026","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"2 September 2026","order":2,"name":"revised","label":"Revised","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"17 September 2026","order":3,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"13 October 2026","order":4,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"This study used publicly available, deidentified radiographs and recorded task-related ratings from study investigators. According to applicable institutional policy, formal institutional review board approval was not required.","order":1,"name":"Ethics","label":"Ethics Approval","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable. This study used publicly available, deidentified chest radiographs and did not involve the recruitment of human participants or the collection of identifiable personal data.","order":2,"name":"Ethics","label":"Consent to Participate","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Not applicable. This manuscript does not contain any individual person\u2019s identifiable data, images, or details requiring consent for publication.","order":3,"name":"Ethics","label":"Consent for Publication","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The authors declare no competing interests.","order":4,"name":"Ethics","label":"Competing interests","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"Three large language models (GPT-5.5, Gemini 3.5 Flash, and Claude Opus 4.8) were evaluated as study subjects in the classification task and are described in the \u201c\n                      \n                      \u201d section. No generative AI tool was used for autonomous content creation or for generating scientific claims in this manuscript; any use was limited to AI-assisted copyediting for language and formatting, with full human accountability for the final text.","order":5,"name":"Ethics","label":"Use of Large Language Models","group":{"name":"EthicsHeading","label":"Declarations"}}]}}