{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T03:46:07Z","timestamp":1782877567608,"version":"3.54.5"},"reference-count":21,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2025,3,5]],"date-time":"2025-03-05T00:00:00Z","timestamp":1741132800000},"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. Artif. Intell."],"abstract":"<jats:sec><jats:title>Objective<\/jats:title><jats:p>To assess the diagnostic performance of the GPT-4 model in comparison to resident physicians in distinguishing between benign and malignant thyroid nodules using ultrasound images.<\/jats:p><\/jats:sec><jats:sec><jats:title>Methods<\/jats:title><jats:p>This study analyzed 1,145 ultrasound images, including 632 malignant and 513 benign nodules. Both the GPT-4 model and two resident physicians independently classified the nodules using ultrasound images. The diagnostic accuracy of the resident physicians was determined by calculating the average of the individual accuracy rates of the two physicians and this was compared with the performance of the GPT-4 model.<\/jats:p><\/jats:sec><jats:sec><jats:title>Results<\/jats:title><jats:p>The GPT-4 model correctly identified 367 out of 632 malignant nodules (58.07%) and 343 out of 513 benign nodules (66.86%). Resident physicians identified 467 malignant (73.89%) and 383 benign nodules (74.66%). There was a statistically significant difference in the classification of malignant nodules (<jats:italic>p<\/jats:italic>\u202f&amp;lt;\u202f0.001) and benign nodules (<jats:italic>p<\/jats:italic>\u202f=\u202f0.048) between the GPT-4 model and residents. GPT-4 performed better for larger nodules (&amp;gt;1\u202fcm) at 65.38%, compared to 53.77% for smaller nodules (\u22641\u202fcm, <jats:italic>p<\/jats:italic>\u202f=\u202f0.004). The AUC for GPT-4 was 0.67, while residents achieved 0.75.<\/jats:p><\/jats:sec><jats:sec><jats:title>Conclusion<\/jats:title><jats:p>The GPT-4 model shows potential in classifying thyroid nodules, but its diagnostic accuracy remains significantly lower than that of resident physicians, particularly for smaller malignant nodules.<\/jats:p><\/jats:sec>","DOI":"10.3389\/frai.2025.1512438","type":"journal-article","created":{"date-parts":[[2025,3,5]],"date-time":"2025-03-05T07:08:30Z","timestamp":1741158510000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":6,"title":["Comparison of the accuracy of GPT-4 and resident physicians in differentiating benign and malignant thyroid nodules"],"prefix":"10.3389","volume":"8","author":[{"given":"Boxiong","family":"Wei","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiumei","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuhong","family":"Shao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiuming","family":"Sun","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Luzeng","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2025,3,5]]},"reference":[{"key":"ref1","doi-asserted-by":"publisher","DOI":"10.1007\/s00330-024-11035-5","article-title":"Assessing GPT-4 multimodal performance in radiological image analysis","author":"Brin","year":"2024","journal-title":"Eur. Radiol."},{"key":"ref2","doi-asserted-by":"publisher","first-page":"546","DOI":"10.1089\/thy.2016.0372","article-title":"A computer-aided diagnosis system using artificial intelligence for the diagnosis and characterization of thyroid nodules on ultrasound: initial clinical assessment","volume":"27","author":"Choi","year":"2017","journal-title":"Thyroid"},{"key":"ref3","doi-asserted-by":"publisher","first-page":"284","DOI":"10.1016\/j.ics.2005.03.187","article-title":"Real-time analysis of ultrasound images using GPU","volume":"1281","author":"Eidheim","year":"2005","journal-title":"Int. Congr. Ser."},{"key":"ref4","doi-asserted-by":"publisher","first-page":"e231362","DOI":"10.1148\/radiol.231362","article-title":"Potential of ChatGPT and GPT-4 for data Mining of Free-Text CT reports on lung Cancer","volume":"308","author":"Fink","year":"2023","journal-title":"Radiology"},{"key":"ref5","doi-asserted-by":"publisher","first-page":"1369","DOI":"10.2174\/1573405618666220422132251","article-title":"A comparison of the performances of artificial intelligence system and radiologists in the ultrasound diagnosis of thyroid nodules","volume":"18","author":"He","year":"2022","journal-title":"Curr. Med. Imaging"},{"key":"ref6","doi-asserted-by":"publisher","first-page":"500","DOI":"10.1038\/s41568-018-0016-5","article-title":"Artificial intelligence in radiology","volume":"18","author":"Hosny","year":"2018","journal-title":"Nat. Rev. Cancer"},{"key":"ref7","article-title":"GPT-4V cannot generate radiology reports yet","author":"Jiang","year":"2024"},{"key":"ref8","doi-asserted-by":"publisher","first-page":"749","DOI":"10.1038\/nrclinonc.2017.141","article-title":"Radiomics: the bridge between medical imaging and personalized medicine","volume":"14","author":"Lambin","year":"2017","journal-title":"Nat. Rev. Clin. Oncol."},{"key":"ref9","doi-asserted-by":"publisher","first-page":"e13746","DOI":"10.1111\/obr.13746","article-title":"ChatGPT\/GPT-4 (large language models): opportunities and challenges of perspective in bariatric healthcare professionals","volume":"25","author":"Law","year":"2024","journal-title":"Obes. Rev."},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1101\/2023.11.03.23298067","article-title":"A comprehensive study of GPT-4V\u2019s multimodal capabilities in medical imaging","author":"Li","year":"2023","journal-title":"medRxiv"},{"key":"ref11","doi-asserted-by":"publisher","first-page":"261","DOI":"10.1016\/j.eng.2018.11.020","article-title":"Deep learning in medical ultrasound analysis: a review","volume":"5","author":"Liu","year":"2019","journal-title":"Engineering"},{"key":"ref12","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2303.13375","article-title":"Capabilities of GPT-4 on medical challenge problems","author":"Nori","year":"2023","journal-title":"CoRR"},{"key":"ref13","doi-asserted-by":"publisher","first-page":"1616","DOI":"10.1007\/s00330-018-5674-x","article-title":"Demystification of AI-driven medical image interpretation: past, present and future","volume":"29","author":"Savadjiev","year":"2019","journal-title":"Eur. Radiol."},{"key":"ref14","doi-asserted-by":"publisher","first-page":"1310","DOI":"10.1002\/jmri.26878","article-title":"Artificial intelligence in the interpretation of breast cancer on MRI","volume":"51","author":"Sheth","year":"2020","journal-title":"J. Magn. Reson. Imaging"},{"key":"ref15","article-title":"Toward parallel optimal computation of ultrasound computed tomography using GPU","author":"Sun","year":"2018"},{"key":"ref16","doi-asserted-by":"publisher","first-page":"1018321","DOI":"10.3389\/fendo.2022.1018321","article-title":"Identification of benign and malignant thyroid nodules based on dynamic AI ultrasound intelligent auxiliary diagnosis system","volume":"13","author":"Wang","year":"2022","journal-title":"Front. Endocrinol. (Lausanne)"},{"key":"ref17","doi-asserted-by":"publisher","first-page":"1602","DOI":"10.21037\/qims-23-1180","article-title":"Assessing the role of GPT-4 in thyroid ultrasound diagnosis and treatment recommendations: enhancing interpretability with a chain of thought approach","volume":"14","author":"Wang","year":"2024","journal-title":"Quant. Imaging Med. Surg."},{"key":"ref18","doi-asserted-by":"publisher","first-page":"112","DOI":"10.1148\/radiol.2019182128","article-title":"Using artificial intelligence to revise ACR TI-RADS risk stratification of thyroid nodules: diagnostic accuracy and utility","volume":"292","author":"Wildman-Tobriner","year":"2019","journal-title":"Radiology"},{"key":"ref19","doi-asserted-by":"publisher","first-page":"4","DOI":"10.1148\/radiol.2020192224","article-title":"Preparing medical imaging data for machine learning","volume":"295","author":"Willemink","year":"2020","journal-title":"Radiology"},{"key":"ref20","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2310.09909","article-title":"Can GPT-4V(ision) serve medical applications? Case studies on GPT-4V for multimodal medical diagnosis","author":"Wu","year":"2023","journal-title":"CoRR"},{"key":"ref21","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2302.08722","article-title":"GPT4MIA: Utilizing Geneative pre-trained transformer (GPT-3) as a plug-and-play Transductive model for medical image analysis","author":"Zhang","year":"2023","journal-title":"arXiv"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1512438\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,3,5]],"date-time":"2025-03-05T07:08:33Z","timestamp":1741158513000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2025.1512438\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,3,5]]},"references-count":21,"alternative-id":["10.3389\/frai.2025.1512438"],"URL":"https:\/\/doi.org\/10.3389\/frai.2025.1512438","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,3,5]]},"article-number":"1512438"}}