{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,11,11]],"date-time":"2026-11-11T05:51:03Z","timestamp":1794376263777,"version":"4.9.0"},"reference-count":17,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T00:00:00Z","timestamp":1776816000000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T00:00:00Z","timestamp":1776816000000},"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":["Insights Imaging"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Objectives<\/jats:title>\n                    <jats:p>Coronary computed tomography angiography (CCTA) has become a cornerstone in non-invasive CAD diagnosis and risk stratification. To standardize reporting and improve clinical decision-making, the CAD-RADS 2.0 system was introduced. This study evaluates the performance of four LLMs, GPT-4o, Gemini 2.0 Flash, DeepSeek V, and Copilot in generating CAD-RADS 2.0-compliant conclusions from standardized CCTA reports.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Materials and methods<\/jats:title>\n                    <jats:p>A total of 196 anonymized CCTA reports were retrospectively analyzed. Each LLM was prompted to provide CAD-RADS 2.0 classifications and follow-up recommendations. Ground truth labels were assigned by a senior radiologist. Performance metrics (accuracy, precision, recall, F1-score), execution times, and agreement (Cohen\u2019s kappa) with expert interpretation were computed. Interobserver agreement between junior and senior radiologists was also assessed.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      LLMs demonstrated good-to-excellent agreement with expert classifications: DeepSeek V (\u03ba\u2009=\u20090.771), Copilot (\u03ba\u2009=\u20090.761), GPT-4o (\u03ba\u2009=\u20090.759), and Gemini 2.0 Flash (\u03ba\u2009=\u20090.634). DeepSeek V achieved the highest accuracy (91.83%). Intra-model consistency was perfect (\u03ba\u2009=\u20091). However, LLMs failed to assign CAD-RADS modifiers. ChatGPT-4o provided the most accurate follow-up recommendations (71.94%). All LLMs outperformed radiologists in execution time (3\u20139\u2009s vs. 15\u201320\u2009s;\n                      <jats:italic>p<\/jats:italic>\n                      \u2009&lt;\u20090.05).\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusions<\/jats:title>\n                    <jats:p>Generic LLMs demonstrate promising performance in automating CAD-RADS 2.0 classification from CCTA reports. However, limitations in modifier assignment and recommendation accuracy highlight areas for refinement before clinical integration.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Critical relevance statement<\/jats:title>\n                    <jats:p>This study explores the potential of large language models to facilitate standardized CAD-RADS 2.0 reporting from coronary CT angiography, highlighting a possible avenue to support workflow efficiency and clinical decision-making in non-invasive coronary artery disease evaluation.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Key Points<\/jats:title>\n                    <jats:p>\n                      <jats:list list-type=\"bullet\">\n                        <jats:list-item>\n                          <jats:p>LLMs demonstrated strong potential in automating CAD-RADS 2.0-compliant structured reporting for CCTA.<\/jats:p>\n                        <\/jats:list-item>\n                        <jats:list-item>\n                          <jats:p>LLMs could significantly enhance efficiency in radiological reporting.<\/jats:p>\n                        <\/jats:list-item>\n                        <jats:list-item>\n                          <jats:p>LLMs need further optimization before clinical integration.<\/jats:p>\n                        <\/jats:list-item>\n                      <\/jats:list>\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Graphical Abstract<\/jats:title>\n                  <\/jats:sec>","DOI":"10.1186\/s13244-026-02285-6","type":"journal-article","created":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T03:47:47Z","timestamp":1776829667000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":4,"title":["Comparative evaluation of large language models for generating CAD-RADS 2.0-compliant diagnostic conclusions in cardiac CT reports"],"prefix":"10.1186","volume":"17","author":[{"given":"Giovanni","family":"Lorusso","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Giorgio","family":"Ruscino","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0009-5882-5316","authenticated-orcid":false,"given":"Alessia","family":"Spitaleri","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Chiara","family":"Morelli","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sara","family":"Greco","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ilaria","family":"Villanova","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicola Maria","family":"Lucarelli","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michele","family":"Mariano","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Amato Antonio","family":"Stabile Ianora","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Nicola","family":"Maggialetti","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,4,22]]},"reference":[{"key":"2285_CR1","doi-asserted-by":"publisher","first-page":"2320","DOI":"10.1016\/S0735-1097(24)04310-9","volume":"83","author":"B Stark","year":"2024","unstructured":"Stark B, Johnson C, Roth GA (2024) Global prevalence of coronary artery disease: an update from the Global Burden of Disease Study. J Am Coll Cardiol 83:2320","journal-title":"J Am Coll Cardiol"},{"key":"2285_CR2","doi-asserted-by":"publisher","first-page":"409","DOI":"10.1714\/3190.31684","volume":"20","author":"MT Mallus","year":"2019","unstructured":"Mallus MT, Prati F (2019) Utilizzo appropriato dell\u2019angiografia coronarica con tomografia computerizzata nella cardiopatia ischemica: indicazioni e limiti. G Ital Cardiol 20:409\u2013416. https:\/\/doi.org\/10.1714\/3190.31684","journal-title":"G Ital Cardiol"},{"key":"2285_CR3","doi-asserted-by":"publisher","first-page":"1226","DOI":"10.1016\/j.jacc.2020.06.076","volume":"76","author":"KM Abdelrahman","year":"2020","unstructured":"Abdelrahman KM, Chen MY, Dey AK et al (2020) Coronary computed tomography angiography from clinical uses to emerging technologies: JACC state-of-the-art review. J Am Coll Cardiol 76:1226\u20131243","journal-title":"J Am Coll Cardiol"},{"key":"2285_CR4","doi-asserted-by":"publisher","first-page":"1458","DOI":"10.1016\/j.jacr.2016.04.024","volume":"13","author":"RC Cury","year":"2016","unstructured":"Cury RC, Abbara S, Achenbach S et al (2016) CAD-RADS\u2122: coronary artery disease\u2013reporting and data system: an expert consensus document of the Society of Cardiovascular Computed Tomography (SCCT), the American College of Radiology (ACR) and the North American Society for Cardiovascular Imaging (NASCI). Endorsed by the American College of Cardiology. J Am Coll Radiol 13:1458\u20131466.e9. https:\/\/doi.org\/10.1016\/j.jacr.2016.04.024","journal-title":"J Am Coll Radiol"},{"key":"2285_CR5","doi-asserted-by":"publisher","first-page":"1974","DOI":"10.1016\/j.jcmg.2022.07.002","volume":"15","author":"RC Cury","year":"2022","unstructured":"Cury RC, Leipsic J, Abbara S et al (2022) CAD-RADS\u2122 2.0\u20142022 coronary artery disease-reporting and data system: an expert consensus document of the Society of Cardiovascular Computed Tomography (SCCT), the American College of Cardiology (ACC), the American College of Radiology (ACR), and the North America Society of Cardiovascular Imaging (NASCI). JACC Cardiovasc Imaging 15:1974\u20132001. https:\/\/doi.org\/10.1016\/j.jcmg.2022.07.002","journal-title":"JACC Cardiovasc Imaging"},{"key":"2285_CR6","doi-asserted-by":"publisher","first-page":"78","DOI":"10.1016\/j.jcmg.2017.08.026","volume":"11","author":"JX Xie","year":"2018","unstructured":"Xie JX, Cury RC, Leipsic J et al (2018) The coronary artery disease-reporting and data system (CAD-RADS): prognostic and clinical implications associated with standardized coronary computed tomography angiography reporting. JACC Cardiovasc Imaging 11:78\u201389. https:\/\/doi.org\/10.1016\/j.jcmg.2017.08.026","journal-title":"JACC Cardiovasc Imaging"},{"key":"2285_CR7","doi-asserted-by":"publisher","first-page":"1849","DOI":"10.1007\/s11547-024-01915-7","volume":"129","author":"M Parillo","year":"2024","unstructured":"Parillo M, Vaccarino F, Beomonte Zobel B, Mallio CA (2024) ChatGPT and radiology report: potential applications and limitations. Radiol Med 129:1849\u20131863. https:\/\/doi.org\/10.1007\/s11547-024-01915-7","journal-title":"Radiol Med"},{"key":"2285_CR8","doi-asserted-by":"publisher","first-page":"755","DOI":"10.1007\/s11547-023-01634-5","volume":"128","author":"E Neri","year":"2023","unstructured":"Neri E, Aghakhanyan G, Zerunian M et al (2023) Explainable AI in radiology: a white paper of the Italian Society of Medical and Interventional Radiology. Radiol Med 128:755\u2013764. https:\/\/doi.org\/10.1007\/s11547-023-01634-5","journal-title":"Radiol Med"},{"key":"2285_CR9","doi-asserted-by":"crossref","unstructured":"Saliba T, Ferrari J, Pozzessere C, Rotzinger D, Fahrni G (2025) Can advanced large language models support radiology training? A performance assessment of DeepSeek R1. Eur J Radiol Artif Intell 3:100024","DOI":"10.1016\/j.ejrai.2025.100024"},{"key":"2285_CR10","doi-asserted-by":"publisher","first-page":"1564","DOI":"10.1007\/s00261-024-04597-x","volume":"50","author":"J Arribas Anta","year":"2025","unstructured":"Arribas Anta J, Moreno-Vedia J, Garc\u00eda L\u00f3pez J et al (2025) Artificial intelligence for detection and characterization of focal hepatic lesions: a review. Abdom Radiol 50:1564\u20131583. https:\/\/doi.org\/10.1007\/s00261-024-04597-x","journal-title":"Abdom Radiol"},{"key":"2285_CR11","doi-asserted-by":"publisher","first-page":"e232756","DOI":"10.1148\/radiol.232756","volume":"310","author":"R Bhayana","year":"2024","unstructured":"Bhayana R (2024) Chatbots and large language models in radiology: a practical primer for clinical and research applications. Radiology 310:e232756. https:\/\/doi.org\/10.1148\/radiol.232756","journal-title":"Radiology"},{"key":"2285_CR12","doi-asserted-by":"publisher","unstructured":"Kotkar AD, Mahadik RS, More PG, Thorat SA (2024) Comparative analysis of transformer-based large language models (LLMs) for text summarization. In: Proceedings of the 2024 1st international conference on advanced computing and emerging technologies (ACET), Ghaziabad, pp 1\u20137, https:\/\/doi.org\/10.1109\/ACET61898.2024.10730348","DOI":"10.1109\/ACET61898.2024.10730348"},{"key":"2285_CR13","doi-asserted-by":"publisher","unstructured":"Silbergleit M, T\u00f3th A, Chamberlin JH et al (2024) ChatGPT vs Gemini: comparative accuracy and efficiency in CAD-RADS score assignment from radiology reports. J Imaging Inform Med. https:\/\/doi.org\/10.1007\/s10278-024-01328-y","DOI":"10.1007\/s10278-024-01328-y"},{"key":"2285_CR14","doi-asserted-by":"publisher","first-page":"464","DOI":"10.1007\/978-3-031-45673-2_46","volume":"14348","author":"Z, Zhong A, Li Y Liu","year":"2023","unstructured":"Liu Z, Zhong A, Li Y et al (2023) Tailoring large language models to radiology: a preliminary approach to LLM adaptation for a highly specialized domain. Mach Learn Med Imaging 14348:464\u2013473","journal-title":"Mach Learn Med Imaging"},{"key":"2285_CR15","doi-asserted-by":"publisher","first-page":"808","DOI":"10.1007\/s11547-023-01651-4","volume":"128","author":"CA Mallio","year":"2023","unstructured":"Mallio CA, Sertorio AC, Bernetti C, Beomonte Zobel B (2023) Large language models for structured reporting in radiology: performance of GPT-4, ChatGPT-3.5, Perplexity and Bing. Radiol Med 128:808\u2013812. https:\/\/doi.org\/10.1007\/s11547-023-01651-4","journal-title":"Radiol Med"},{"key":"2285_CR16","doi-asserted-by":"publisher","unstructured":"Rahman A, Mahir SH, Tashrif MTA et al (2025) Comparative analysis based on DeepSeek, ChatGPT, and Google Gemini: Features, techniques, performance, future prospects. Syst Soft Comput 7:200396. https:\/\/doi.org\/10.1016\/j.sasc.2025.200396","DOI":"10.1016\/j.sasc.2025.200396"},{"key":"2285_CR17","doi-asserted-by":"publisher","first-page":"1445","DOI":"10.1007\/s11547-023-01711-9","volume":"128","author":"CA Mallio","year":"2023","unstructured":"Mallio CA, Bernetti C, Sertorio AC, Beomonte Zobel B (2023) Large language models and structured reporting: never stop chasing critical thinking. Radiol Med 128:1445\u20131446. https:\/\/doi.org\/10.1007\/s11547-023-01711-9","journal-title":"Radiol Med"}],"container-title":["Insights into Imaging"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13244-026-02285-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1186\/s13244-026-02285-6","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1186\/s13244-026-02285-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,22]],"date-time":"2026-04-22T03:47:51Z","timestamp":1776829671000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1186\/s13244-026-02285-6"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,22]]},"references-count":17,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["2285"],"URL":"https:\/\/doi.org\/10.1186\/s13244-026-02285-6","relation":{},"ISSN":["1869-4101"],"issn-type":[{"value":"1869-4101","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,22]]},"assertion":[{"value":"28 September 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"21 March 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"22 April 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The present retrospective study was conducted at the University of Bari (Italy), in full accordance with ethical principles, including the World Medical Association Declaration of Helsinki and the additional requirements of Italian law. Furthermore, the University of Bari, Italy, classified the study as being exempt from ethical review as it carries only negligible risk and involves the use of existing data that contain only non-identifiable information about human beings. The patients signed a written informed consent form.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethics approval and consent to participate"}},{"value":"Informed consent was obtained from all subjects involved in the study.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Consent for publication"}},{"value":"The authors declare that they have no competing interests.","order":4,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}}],"article-number":"112"}}