{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T05:18:17Z","timestamp":1787894297598,"version":"build-2784847793"},"reference-count":27,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T00:00:00Z","timestamp":1787875200000},"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. Med."],"abstract":"<jats:sec>\n                    <jats:title>Objective<\/jats:title>\n                    <jats:p>To compare reasoning vs. conventional large language models (LLMs) in generating answers with guideline-aligned explanations for Breast Imaging Reporting and Data System (BI-RADS) educational questions.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>In this prospective study performed from February 6 to 12, 2025, 49 English-Chinese question pairs were extracted from BI-RADS Atlas Fifth Edition. Two reasoning LLMs (ChatGPT-o1, Deepseek-R1) and six conventional LLMs (Gemini2.0-Flash, Deepseek-V3, ChatGPT-4o, ChatGPT-3.5, Qwen-2.5, and WenXinYiYan-3.5) generated answers and explanations to the questions through structured prompts. Three radiologists specialized in breast imaging independently evaluated responses using a 5-point Likert scale, with reference to standard answers.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      The reasoning LLMs significantly outperformed conventional models (median [interquartile range (IQR)]: 3.7 [2.7\u20134.0] vs. 2.7 [2.0\u20133.7],\n                      <jats:italic>P<\/jats:italic>\n                      \u2009&amp;lt;\u20090.001), with ChatGPT-o1 and Deepseek-R1 demonstrating peak performance. Both categories of LLMs exhibited significant score reductions in handling questions with multifaceted clinical scenarios (reasoning models: median 4.0 [2.7\u20134.3] vs. 2.7 [2.3\u20132.7],\n                      <jats:italic>\u0394<\/jats:italic>\n                      median\u2009=\u2009\u22121.3,\n                      <jats:italic>P<\/jats:italic>\n                      \u2009&amp;lt;\u20090.001; conventional models: 2.7 [2.0\u20133.7] vs. 2.3 [2.0\u20132.7],\n                      <jats:italic>\u0394<\/jats:italic>\n                      median\u2009=\u2009\u22120.4,\n                      <jats:italic>P<\/jats:italic>\n                      \u2009&amp;lt;\u20090.001). While question language showed no significant impact on reasoning LLMs (ChatGPT-o1 and Deepseek-R1), it affected some conventional models (ChatGPT-3.5, Deepseek-V3 and Gemini2.0-Flash). LLMs performance remained independent of question section and question type.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Reasoning LLMs show significant potential for BI-RADS guideline explanation and education, but require specific optimization for complex clinical scenario instruction.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.3389\/fmed.2026.1926969","type":"journal-article","created":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T04:33:33Z","timestamp":1787891613000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Reasoning vs. conventional large language models for BI-RADS educational questions answering: a multi-model comparative evaluation"],"prefix":"10.3389","volume":"13","author":[{"given":"Yuxia","family":"Tang","sequence":"first","affiliation":[{"name":"Department of Radiology, The First Affiliated Hospital with Nanjing Medical University","place":["Nanjing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yuting","family":"Liu","sequence":"additional","affiliation":[{"name":"Department of Radiology, The First Affiliated Hospital with Nanjing Medical University","place":["Nanjing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Mengxuan","family":"Liu","sequence":"additional","affiliation":[{"name":"Department of Radiology, The First Affiliated Hospital with Nanjing Medical University","place":["Nanjing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hao","family":"Ni","sequence":"additional","affiliation":[{"name":"Department of Mathematics, University College London","place":["London, United Kingdom"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Siqi","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Radiology, The First Affiliated Hospital with Nanjing Medical University","place":["Nanjing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shouju","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Radiology, The First Affiliated Hospital with Nanjing Medical University","place":["Nanjing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1965","published-online":{"date-parts":[[2026,8,28]]},"reference":[{"key":"B1","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":"B2","doi-asserted-by":"publisher","first-page":"1930","DOI":"10.1038\/s41591-023-02448-8","article-title":"Large language models in medicine","volume":"29","author":"Thirunavukarasu","year":"2023","journal-title":"Nat Med"},{"key":"B3","doi-asserted-by":"publisher","first-page":"e240701","DOI":"10.1148\/radiol.240701","article-title":"Large-Scale validation of the feasibility of GPT-4 as a proofreading tool for head CT reports","volume":"314","author":"Kim","year":"2025","journal-title":"Radiology"},{"key":"B4","doi-asserted-by":"publisher","first-page":"e50945","DOI":"10.2196\/50945","article-title":"The role of large language models in medical education: applications and implications","volume":"9","author":"Safranek","year":"2023","journal-title":"JMIR Med Educ"},{"key":"B5","doi-asserted-by":"publisher","first-page":"320","DOI":"10.1080\/0142159X.2023.2290997","article-title":"The tools of the future are the challenges of today: the use of ChatGPT in problem-based learning medical education","volume":"46","author":"Divito","year":"2024","journal-title":"Med Teach"},{"key":"B6","doi-asserted-by":"publisher","first-page":"e46885","DOI":"10.2196\/46885","article-title":"The role of ChatGPT, generative language models, and artificial intelligence in medical education: a conversation with ChatGPT and a call for papers","volume":"9","author":"Eysenbach","year":"2023","journal-title":"JMIR Med Educ"},{"key":"B7","doi-asserted-by":"publisher","first-page":"e0000198","DOI":"10.1371\/journal.pdig.0000198","article-title":"Performance of ChatGPT on USMLE: potential for AI-assisted medical education using large language models","volume":"2","author":"Kung","year":"2023","journal-title":"PLOS Digit Health"},{"key":"B8","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":"B9","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2502.10867","article-title":"A tutorial on LLM reasoning: relevant methods behind ChatGPT o1","author":"Wang","year":"2025","journal-title":"arXiv"},{"key":"B10","doi-asserted-by":"publisher","first-page":"633","DOI":"10.1038\/s41586-025-09422-z","article-title":"DeepSeek-R1: incentivizing reasoning capability in LLMs via reinforcement learning","volume":"645","author":"Guo","year":"2025","journal-title":"Nature"},{"key":"B11","doi-asserted-by":"publisher","first-page":"2546","DOI":"10.1038\/s41591-025-03727-2","article-title":"Benchmark evaluation of DeepSeek large language models in clinical decision-making","volume":"31","author":"Sandmann","year":"2025","journal-title":"Nat Med"},{"key":"B12","doi-asserted-by":"publisher","first-page":"2550","DOI":"10.1038\/s41591-025-03726-3","article-title":"Comparative benchmarking of the DeepSeek large language model on medical tasks and clinical reasoning","volume":"31","author":"Tordjman","year":"2025","journal-title":"Nat Med"},{"key":"B13","volume-title":"Breast Imaging Reporting and Data System","author":"Magny","year":"2025"},{"key":"B14","doi-asserted-by":"publisher","first-page":"871","DOI":"10.1148\/radiol.2243011626","article-title":"Does training in the breast imaging reporting and data system (BI-RADS) improve biopsy recommendations or feature analysis agreement with experienced breast imagers at mammography?","volume":"224","author":"Berg","year":"2002","journal-title":"Radiology"},{"key":"B15","doi-asserted-by":"publisher","first-page":"845","DOI":"10.1148\/radiology.211.3.r99jn31845","article-title":"BI-RADS categorization as a predictor of malignancy","volume":"211","author":"Orel","year":"1999","journal-title":"Radiology"},{"key":"B16","doi-asserted-by":"publisher","first-page":"4468","DOI":"10.1007\/s00330-018-5934-9","article-title":"Effect of training on ultrasonography (US) BI-RADS features for radiology residents: a multicenter study comparing performances after training","volume":"29","author":"Yoon","year":"2019","journal-title":"Eur Radiol"},{"key":"B17","doi-asserted-by":"publisher","first-page":"862","DOI":"10.1148\/radiol.2273020046","article-title":"Survey of radiology residents: breast imaging training and attitudes","volume":"227","author":"Bassett","year":"2003","journal-title":"Radiology"},{"key":"B18","doi-asserted-by":"publisher","first-page":"263","DOI":"10.2214\/AJR.10.5834","article-title":"Breast imaging training and attitudes: update survey of senior radiology residents","volume":"197","author":"Bassett","year":"2011","journal-title":"AJR Am J Roentgenol"},{"key":"B19","doi-asserted-by":"publisher","first-page":"246","DOI":"10.1016\/j.jacr.2008.11.014","article-title":"Female medical students\u2019 interest in radiology careers","volume":"6","author":"Roubidoux","year":"2009","journal-title":"J Am Coll Radiol"},{"key":"B20","doi-asserted-by":"publisher","first-page":"327","DOI":"10.2214\/AJR.07.3346","article-title":"Predictors of radiologists\u2019 perceived risk of malpractice lawsuits in breast imaging","volume":"192","author":"Dick","year":"2009","journal-title":"AJR Am J Roentgenol"},{"key":"B21","doi-asserted-by":"publisher","first-page":"370","DOI":"10.2214\/AJR.08.1665","article-title":"Workforce shortages in breast imaging: impact on mammography utilization","volume":"192","author":"Wing","year":"2009","journal-title":"AJR Am J Roentgenol"},{"key":"B22","doi-asserted-by":"publisher","first-page":"894","DOI":"10.1016\/j.jacr.2014.02.005","article-title":"A survey of breast imaging fellowship programs: current status of curriculum and training in the United States and Canada","volume":"11","author":"Farria","year":"2014","journal-title":"J Am Coll Radiol"},{"key":"B23","doi-asserted-by":"publisher","first-page":"888","DOI":"10.1016\/j.acra.2014.01.025","article-title":"Radiology resident mammography training: interpretation difficulty and error-making patterns","volume":"21","author":"Grimm","year":"2014","journal-title":"Acad Radiol"},{"key":"B24","doi-asserted-by":"publisher","first-page":"26","DOI":"10.1016\/j.clinimag.2019.12.001","article-title":"Breast cancer malpractice litigation: a 10-year analysis and update in trends","volume":"60","author":"Lee","year":"2020","journal-title":"Clin Imaging"},{"key":"B25","doi-asserted-by":"publisher","first-page":"e2346721","DOI":"10.1001\/jamanetworkopen.2023.46721","article-title":"Performance of large language models on a neurology board-style examination","volume":"6","author":"Schubert","year":"2023","journal-title":"JAMA Netw Open"},{"key":"B26","doi-asserted-by":"publisher","first-page":"e2417641","DOI":"10.1001\/jamanetworkopen.2024.17641","article-title":"Performance of large language models on medical oncology examination questions","volume":"7","author":"Longwell","year":"2024","journal-title":"JAMA Netw Open"},{"key":"B27","doi-asserted-by":"publisher","first-page":"1616145","DOI":"10.3389\/frai.2025.1616145","article-title":"Medical reasoning in LLMs: an in-depth analysis of DeepSeek R1","volume":"8","author":"Moell","year":"2025","journal-title":"Front Artif Intell"}],"container-title":["Frontiers in Medicine"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1926969\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,8,28]],"date-time":"2026-08-28T04:33:44Z","timestamp":1787891624000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/fmed.2026.1926969\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,8,28]]},"references-count":27,"alternative-id":["10.3389\/fmed.2026.1926969"],"URL":"https:\/\/doi.org\/10.3389\/fmed.2026.1926969","relation":{},"ISSN":["2296-858X"],"issn-type":[{"value":"2296-858X","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,8,28]]},"article-number":"1926969"}}