{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,6,25]],"date-time":"2026-06-25T15:08:03Z","timestamp":1782400083319,"version":"3.54.5"},"reference-count":29,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T00:00:00Z","timestamp":1772582400000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Digit. Health"],"abstract":"<jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>To evaluate three large language models (LLMs), including ChatGPT 5, ChatGPT 4o, and ChatGPT 3.5, in automating TNM staging from PET-CT reports across six cancer types, and to assess their clinical utility compared with junior radiologists.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Materials and methods<\/jats:title>\n                    <jats:p>\n                      PET-CT reports from 552 treatment-naive patients in two institutions with confirmed primary malignancies (lung, breast, liver, pancreatic, renal, and prostate cancer) were analyzed. Three ChatGPT-series LLMs and five junior radiologists independently performed TNM staging. Reference standards were established by two senior radiologists according to the 8th version of American Joint Committee on Cancer (AJCC) staging system. Performance was evaluated using accuracy rates. Intra-model agreement was assessed by repeating each model three times per report with identical prompts, and inter-model agreement was evaluated using Cohen's\n                      <jats:italic>\u03ba<\/jats:italic>\n                      coefficients.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      ChatGPT 5 achieved the highest overall accuracy (82.1%, 453\/552), followed by ChatGPT 4o (74.3%, 410\/552), both significantly outperforming ChatGPT 3.5 (59.6%, 329\/552) and junior radiologists (77.0%, 425\/552;\n                      <jats:italic>p<\/jats:italic>\n                      \u2009=\u20090.041 for ChatGPT 5 vs. junior radiologists). Accuracy varied by cancer type, with the highest performance in lung cancer staging (88.5%) and the lowest in pancreatic cancer (69.2%). Across TNM categories, all models achieved the best performance in T staging, followed by N staging, with M staging remaining the most challenging. ChatGPT 5 showed near-perfect intra-model agreement (\n                      <jats:italic>\u03ba<\/jats:italic>\n                      \u2009=\u20090.96), while inter-model agreement ranged from moderate between ChatGPT 3.5 and 4o (\n                      <jats:italic>\u03ba<\/jats:italic>\n                      \u2009=\u20090.58) to substantial between ChatGPT 5 and 4o (\n                      <jats:italic>\u03ba<\/jats:italic>\n                      \u2009=\u20090.78). ChatGPT 5 processed cases markedly faster than junior radiologists (8.3\u2009\u00b1\u20093.2 vs. 92.5\u2009\u00b1\u200921.7\u2005s per case;\n                      <jats:italic>p<\/jats:italic>\n                      \u2009&amp;lt;\u20090.001).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>Among the three LLMs, ChatGPT 5 demonstrated the highest accuracy, stability, and efficiency in automated TNM staging from PET-CT reports, achieving performance comparable to or slightly exceeding junior radiologists. Its advantages in T staging and lung cancer evaluation highlight its clinical utility as a potential decision-support tool.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fdgth.2026.1741973","type":"journal-article","created":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T07:02:54Z","timestamp":1772607774000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":1,"title":["Evaluating large language models for automated TNM staging from PET-CT reports: a multi-cancer comparative study"],"prefix":"10.3389","volume":"8","author":[{"given":"Wen","family":"Xu","sequence":"first","affiliation":[{"name":"Department of Radiology, Hangzhou First People\u2019s Hospital","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Lixiu","family":"Cao","sequence":"additional","affiliation":[{"name":"Department of Nuclear Medicine Imaging, Tangshan People\u2019s Hospital","place":["Tangshan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Qijun","family":"Shen","sequence":"additional","affiliation":[{"name":"Department of Radiology, Hangzhou First People\u2019s Hospital","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yanna","family":"Shan","sequence":"additional","affiliation":[{"name":"Department of Radiology, Hangzhou First People\u2019s Hospital","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shushu","family":"Pan","sequence":"additional","affiliation":[{"name":"Department of Radiology, Hangzhou First People\u2019s Hospital","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mei","family":"Ruan","sequence":"additional","affiliation":[{"name":"Department of Radiology, Hangzhou First People\u2019s Hospital","place":["Hangzhou, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,3,4]]},"reference":[{"key":"B1","doi-asserted-by":"publisher","first-page":"93","DOI":"10.3322\/caac.21388","article-title":"The eighth edition AJCC cancer staging manual: continuing to build a bridge from a population-based to a more \u201cpersonalized\u201d approach to cancer staging","volume":"67","author":"Amin","year":"2017","journal-title":"CA Cancer J Clin"},{"key":"B2","doi-asserted-by":"publisher","first-page":"126","DOI":"10.3348\/kjr.2023.0997","article-title":"Large language models: a guide for radiologists","volume":"25","author":"Kim","year":"2024","journal-title":"Korean J Radiol"},{"key":"B3","doi-asserted-by":"publisher","first-page":"e230725","DOI":"10.1148\/radiol.230725","article-title":"Leveraging GPT-4 for post hoc transformation of free-text radiology reports into structured reporting: a multilingual feasibility study","volume":"307","author":"Adams","year":"2023","journal-title":"Radiology"},{"key":"B4","doi-asserted-by":"publisher","first-page":"e231259","DOI":"10.1148\/radiol.231259","article-title":"Evaluating GPT-4 on impressions generation in radiology reports","volume":"307","author":"Sun","year":"2023","journal-title":"Radiology"},{"key":"B5","doi-asserted-by":"publisher","first-page":"e232133","DOI":"10.1148\/radiol.232133","article-title":"BI-RADS category assignments by GPT-3.5, GPT-4, and google bard: a multilanguage study","volume":"311","author":"Cozzi","year":"2024","journal-title":"Radiology"},{"key":"B6","doi-asserted-by":"publisher","first-page":"e230582","DOI":"10.1148\/radiol.230582","article-title":"Performance of ChatGPT on a radiology board-style examination: insights into current strengths and limitations","volume":"307","author":"Bhayana","year":"2023","journal-title":"Radiology"},{"key":"B7","doi-asserted-by":"publisher","first-page":"e244630","DOI":"10.1001\/jamanetworkopen.2024.4630","article-title":"Quality of large language model responses to radiation oncology patient care questions","volume":"7","author":"Yalamanchili","year":"2024","journal-title":"JAMA Netw Open"},{"key":"B8","doi-asserted-by":"publisher","first-page":"141","DOI":"10.1038\/s43856-023-00370-1","article-title":"The future landscape of large language models in medicine","volume":"3","author":"Clusmann","year":"2023","journal-title":"Commun Med"},{"key":"B9","doi-asserted-by":"publisher","first-page":"e48659","DOI":"10.2196\/48659","article-title":"Assessing the utility of ChatGPT throughout the entire clinical workflow: development and usability study","volume":"25","author":"Rao","year":"2023","journal-title":"J Med Internet Res"},{"key":"B10","doi-asserted-by":"publisher","first-page":"172","DOI":"10.1038\/s41586-023-06291-2","article-title":"Large language models encode clinical knowledge","volume":"620","author":"Singhal","year":"2023","journal-title":"Nature"},{"key":"B11","doi-asserted-by":"publisher","first-page":"e231040","DOI":"10.1148\/radiol.231040","article-title":"Diagnostic performance of ChatGPT from patient history and imaging findings on the diagnosis please quizzes","volume":"308","author":"Ueda","year":"2023","journal-title":"Radiology"},{"key":"B12","doi-asserted-by":"publisher","first-page":"e232714","DOI":"10.1148\/radiol.232714","article-title":"Potential of GPT-4 for detecting errors in radiology reports: implications for reporting accuracy","volume":"311","author":"Gertz","year":"2024","journal-title":"Radiology"},{"key":"B13","doi-asserted-by":"publisher","first-page":"2817","DOI":"10.1007\/s00330-023-10213-1","article-title":"ChatGPT makes medicine easy to swallow: an exploratory case study on simplified radiology reports","volume":"34","author":"Jeblick","year":"2024","journal-title":"Eur Radiol"},{"key":"B14","doi-asserted-by":"publisher","first-page":"101068","DOI":"10.1016\/j.jocmr.2024.101068","article-title":"Generative Pre-trained transformer 4 analysis of cardiovascular magnetic resonance reports in suspected myocarditis: a multicenter study","volume":"26","author":"Kaya","year":"2024","journal-title":"J Cardiovasc Magn Reson"},{"key":"B15","doi-asserted-by":"publisher","first-page":"e43958","DOI":"10.7759\/cureus.43958","article-title":"Artificial intelligence (AI) in radiology: a deep dive into ChatGPT 4.0\u2019s accuracy with the American journal of neuroradiology\u2019s (AJNR) \u201ccase of the month\u201d","volume":"15","author":"Suthar","year":"2023","journal-title":"Cureus"},{"key":"B16","doi-asserted-by":"publisher","first-page":"62","DOI":"10.1038\/s41746-023-00988-4","article-title":"Generative AI and large language models in health care: pathways to implementation","volume":"7","author":"Raza","year":"2024","journal-title":"NPJ Dig Med"},{"key":"B17","doi-asserted-by":"publisher","first-page":"506","DOI":"10.1177\/02184923231187279","article-title":"Accuracy of nodal staging by 18F-FDG PET\/CT in limited stage small-cell lung cancer","volume":"31","author":"Hockmann","year":"2023"},{"key":"B18","doi-asserted-by":"publisher","first-page":"118","DOI":"10.1038\/s41746-020-00324-0","article-title":"The state of artificial intelligence-based FDA-approved medical devices and algorithms: an online database","volume":"3","author":"Benjamens","year":"2020","journal-title":"NPJ Digit Med"},{"key":"B19","doi-asserted-by":"publisher","first-page":"e200103","DOI":"10.1148\/ryai.2020200103","article-title":"OPTIMAM mammography image database: a large-scale resource of mammography images and clinical data","volume":"3","author":"Halling-Brown","year":"2020","journal-title":"Radiol Artif Intell"},{"key":"B20","doi-asserted-by":"publisher","first-page":"e2431696","DOI":"10.2214\/AJR.24.31696","article-title":"Lung cancer staging using chest CT and FDG PET\/CT free-text reports: comparison among three ChatGPT large-language models and six human readers","volume":"223","author":"Lee","year":"2024","journal-title":"AJR Am J Roentgenol"},{"key":"B21","article-title":"Holistic evaluation of GPT-4V for biomedical imaging","author":"Liu","year":"2023"},{"key":"B22","doi-asserted-by":"publisher","first-page":"e230877","DOI":"10.1148\/radiol.230877","article-title":"GPT-4 for automated determination of radiological study and protocol based on radiology request forms: a feasibility study","volume":"307","author":"Gertz","year":"2023","journal-title":"Radiology"},{"key":"B23","article-title":"Capabilities of Gemini models in medicine","author":"Saab","year":"2024"},{"key":"B24","doi-asserted-by":"publisher","first-page":"104828","DOI":"10.1016\/j.ijmedinf.2022.104828","article-title":"The 2021 landscape of FDA-approved artificial intelligence\/machine learning-enabled medical devices: an analysis of the characteristics and intended use","volume":"165","author":"Zhu","year":"2022","journal-title":"Int J Med Inf"},{"key":"B25","doi-asserted-by":"publisher","first-page":"e210284","DOI":"10.1148\/ryai.210284","article-title":"Toward foundational deep learning models for medical imaging in the new era of transformer networks","volume":"4","author":"Willemink","year":"2022","journal-title":"Radiol Artif Intell"},{"key":"B26","doi-asserted-by":"publisher","first-page":"1256","DOI":"10.1016\/j.acra.2023.08.039","article-title":"Potential applications and impact of ChatGPT in radiology","volume":"31","author":"Bajaj","year":"2024","journal-title":"Acad Radiol"},{"key":"B27","doi-asserted-by":"publisher","first-page":"9","DOI":"10.1186\/s42492-023-00136-5","article-title":"Translating radiology reports into plain language using ChatGPT and GPT-4 with prompt learning: results, limitations, and potential","volume":"6","author":"Lyu","year":"2023","journal-title":"Vis Comput Ind Biomed Art"},{"key":"B28","doi-asserted-by":"publisher","first-page":"e230205","DOI":"10.1148\/ryai.230205","article-title":"Generative large language models for detection of speech recognition errors in radiology reports","volume":"5","author":"Schmidt","year":"2023","journal-title":"Radiol Artif Intell"},{"key":"B29","doi-asserted-by":"publisher","first-page":"e662","DOI":"10.1016\/S2589-7500(24)00124-9","article-title":"A future role for health applications of large language models depends on regulators enforcing safety standards","volume":"6","author":"Freyer","year":"2024","journal-title":"Lancet Digit Health"}],"container-title":["Frontiers in Digital Health"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2026.1741973\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,4]],"date-time":"2026-03-04T07:02:55Z","timestamp":1772607775000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fdgth.2026.1741973\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,3,4]]},"references-count":29,"alternative-id":["10.3389\/fdgth.2026.1741973"],"URL":"https:\/\/doi.org\/10.3389\/fdgth.2026.1741973","relation":{},"ISSN":["2673-253X"],"issn-type":[{"value":"2673-253X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,3,4]]},"article-number":"1741973"}}