{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,15]],"date-time":"2026-10-15T13:45:18Z","timestamp":1792071918346,"version":"4.4.0"},"reference-count":33,"publisher":"Wiley","issue":"9","license":[{"start":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T00:00:00Z","timestamp":1788307200000},"content-version":"vor","delay-in-days":1,"URL":"http:\/\/onlinelibrary.wiley.com\/termsAndConditions#vor"},{"start":{"date-parts":[[2026,9,1]],"date-time":"2026-09-01T00:00:00Z","timestamp":1788220800000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["aapm.onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["Medical Physics"],"published-print":{"date-parts":[[2026,9]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:sec>\n                    <jats:title>Background<\/jats:title>\n                    <jats:p>Consistently generating clinically acceptable plans without human intervention remains a challenge in radiotherapy. Rule\u2010based automation provides deterministic execution, and knowledge\u2010based planning (KBP) provides statistical dose estimation, but both often require manual refinement. Large language models (LLMs) offer clinical reasoning capability, but effective autonomous planning also requires a mechanism to execute complex planning actions within the treatment planning system (TPS).<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Purpose<\/jats:title>\n                    <jats:p>To develop and evaluate PlanningCopilot, an agentic system that utilizes the reasoning capability of LLM and a validated Eclipse Scripting API (ESAPI) optimization module integrating KBP initialization (\u201cPlanAct\u201d) to autonomously generate treatment plans. This study evaluates the system's ability to produce clinically acceptable plans for locally advanced non\u2010small cell lung cancer (LA\u2010NSCLC) and assesses its potential to refine performance by self\u2010learning.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Methods<\/jats:title>\n                    <jats:p>PlanningCopilot was implemented as a multi\u2010agent framework linked to the TPS through PlanAct API. It comprises four specialized GPT\u20104.1 agents that iteratively interact with the TPS: (1) an Evaluator agent that accesses the plan and generates plan quality reports, (2) a Supervisor agent that validates these reports before passing them to a Planner agent, (3) the Planner agent that executes initialization and optimization tasks through PlanAct API and planning guidelines, and (4) an optional Learner agent that synthesizes optimization history into Planner\u2010facing prompt addendums. We retrospectively analyzed 62 patients with conventionally fractionated LA\u2010NSCLC and compared original clinical plans with autonomous plans with and without the Learner agent. Measurement\u2010based patient\u2010specific quality assurance (PSQA) was performed on the first 21 autonomous IMRT plans in planning order.<\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Results<\/jats:title>\n                    <jats:p>\n                      All autonomous plans met clinical dosimetric requirements, including those not achieved in the clinical plans and KBP (RapidPlan) plans. Paired Wilcoxon signed\u2010rank tests showed no significant differences between autonomous and clinical plans for Lungs D\n                      <jats:sub>mean<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.371), Lungs V\n                      <jats:sub>20Gy<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.449), Lungs V\n                      <jats:sub>5Gy<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.309), Heart D\n                      <jats:sub>50%<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.175), Esophagus D\n                      <jats:sub>mean<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.750), Spinal Cord D\n                      <jats:sub>0.03cc<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.422), and Plan D\n                      <jats:sub>0.03cc<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.941). Furthermore, autonomous plans achieved significantly lower Esophagus D\n                      <jats:sub>0.03cc<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.027). Compared with RapidPlan initialization, PlanningCopilot improved multiple dosimetric endpoints, including Lungs D\n                      <jats:sub>mean<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&lt;\u00a00.001), Lungs V\n                      <jats:sub>20Gy<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&lt;\u00a00.001), Lungs V\n                      <jats:sub>5Gy<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.004), Heart D\n                      <jats:sub>50%<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0=\u00a00.037), and Plan D\n                      <jats:sub>0.03cc<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&lt;\u00a00.001), with the cost of higher Esophagus D\n                      <jats:sub>mean<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&lt;\u00a00.001) and Esophagus D\n                      <jats:sub>0.03cc<\/jats:sub>\n                      (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&lt;\u00a00.001). In a subset of 18 cases requiring at least two iterations, applying Learner\u2010derived knowledge reduced required iterations by an average of 11.8% while maintaining comparable plan quality (\n                      <jats:italic>p<\/jats:italic>\n                      \u00a0&gt;\u00a00.05). All 21 autonomous IMRT plans passed measurement\u2010based PSQA.\n                    <\/jats:p>\n                  <\/jats:sec>\n                  <jats:sec>\n                    <jats:title>Conclusion<\/jats:title>\n                    <jats:p>PlanningCopilot enables autonomous generation of clinically acceptable and deliverable treatment plans for LA\u2010NSCLC. It consistently satisfies clinical dosimetric requirements across varying anatomical complexities and improves optimization efficiency through self\u2010learning from prior optimization history.<\/jats:p>\n                  <\/jats:sec>","DOI":"10.1002\/mp.70665","type":"journal-article","created":{"date-parts":[[2026,9,2]],"date-time":"2026-09-02T07:58:20Z","timestamp":1788335900000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["PlanningCopilot: An agentic framework integrating ESAPI modules for autonomous treatment planning in lung radiotherapy"],"prefix":"10.1002","volume":"53","author":[{"given":"Hao","family":"Guo","sequence":"first","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zipai","family":"Wang","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tenzin","family":"Kunkyab","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yang","family":"Lei","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Robert","family":"Samstein","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kenneth E.","family":"Rosenzweig","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ming","family":"Chao","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Tian","family":"Liu","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jiahan","family":"Zhang","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Junyi","family":"Xia","sequence":"additional","affiliation":[{"name":"Department of Radiation Oncology Icahn School of Medicine at Mount Sinai  New York New York USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,9,2]]},"reference":[{"key":"e_1_2_9_2_1","doi-asserted-by":"publisher","DOI":"10.3322\/caac.21660"},{"key":"e_1_2_9_3_1","doi-asserted-by":"publisher","DOI":"10.1016\/S1470\u20102045(14)71207\u20100"},{"key":"e_1_2_9_4_1","doi-asserted-by":"publisher","DOI":"10.1080\/14737140.2020.1760094"},{"key":"e_1_2_9_5_1","doi-asserted-by":"publisher","DOI":"10.1016\/S1053\u20104296(99)80052\u20106"},{"key":"e_1_2_9_6_1","doi-asserted-by":"publisher","DOI":"10.1287\/inte.2021.1095"},{"key":"e_1_2_9_7_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.prro.2011.11.012"},{"key":"e_1_2_9_8_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.prro.2016.04.005"},{"key":"e_1_2_9_9_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.canrad.2021.06.006"},{"key":"e_1_2_9_10_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.prro.2013.11.007"},{"key":"e_1_2_9_11_1","doi-asserted-by":"publisher","DOI":"10.1118\/1.4761864"},{"key":"e_1_2_9_12_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijrobp.2010.11.030"},{"key":"e_1_2_9_13_1","doi-asserted-by":"publisher","DOI":"10.3390\/cancers17223654"},{"key":"e_1_2_9_14_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.18068"},{"key":"e_1_2_9_15_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijrobp.2019.12.034"},{"key":"e_1_2_9_16_1","doi-asserted-by":"publisher","DOI":"10.3390\/jcm15072531"},{"key":"e_1_2_9_17_1","doi-asserted-by":"publisher","DOI":"10.1177\/1533033819873922"},{"key":"e_1_2_9_18_1","unstructured":"YangD WuX XieY et\u00a0al.Zero\u2010shot large language model agents for fully automated radiotherapy treatment planning. arXiv:2510.11754 Preprint posted online October 12 2025. doi:10.48550\/arXiv.2510.11754"},{"key":"e_1_2_9_19_1","doi-asserted-by":"publisher","DOI":"10.1088\/1361\u20106560\/adbff1"},{"key":"e_1_2_9_20_1","unstructured":"NusratH LuoB HallR et\u00a0al.Autonomous radiotherapy treatment planning using DOLA: a privacy\u2010preserving LLM\u2010Based optimization agent. arXiv:2503.17553 Preprint posted online March 21 2025. doi:10.48550\/arXiv.2503.17553"},{"key":"e_1_2_9_21_1","doi-asserted-by":"publisher","DOI":"10.1186\/s13014\u2010025\u201002660\u20105"},{"key":"e_1_2_9_22_1","doi-asserted-by":"publisher","DOI":"10.1088\/1361\u20106560\/adf02c"},{"key":"e_1_2_9_23_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.17416"},{"key":"e_1_2_9_24_1","doi-asserted-by":"publisher","DOI":"10.1002\/acm2.70304"},{"key":"e_1_2_9_25_1","doi-asserted-by":"publisher","DOI":"10.3390\/curroncol29100552"},{"key":"e_1_2_9_26_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ijrobp.2009.06.091"},{"key":"e_1_2_9_27_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.12810"},{"key":"e_1_2_9_28_1","doi-asserted-by":"publisher","DOI":"10.1080\/0284186X.2023.2238882"},{"key":"e_1_2_9_29_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.prro.2024.04.014"},{"key":"e_1_2_9_30_1","doi-asserted-by":"publisher","DOI":"10.1002\/mp.70657"},{"key":"e_1_2_9_31_1","unstructured":"WuQ BansalG ZhangJ et\u00a0al.AutoGen: enabling next\u2010gen LLM applications via multi\u2010agent conversation. arXiv:2308.08155.\u00a0Preprint posted online October 3 2023. doi:10.48550\/arXiv.2308.08155"},{"key":"e_1_2_9_32_1","unstructured":"SorkaM GorenshteinA AranD ShellyS.A multi\u2010agent approach to neurological clinical reasoning. arXiv:2508.14063. Preprint posted online August 10 2025. doi:10.48550\/arXiv.2508.14063"},{"key":"e_1_2_9_33_1","doi-asserted-by":"publisher","DOI":"10.53555\/hr4af222"},{"key":"e_1_2_9_34_1","doi-asserted-by":"publisher","DOI":"10.1186\/1748\u2010717X\u20106\u2010140"}],"container-title":["Medical Physics"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.70665","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/mp.70665","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/mp.70665","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,10,15]],"date-time":"2026-10-15T12:54:31Z","timestamp":1792068871000},"score":1,"resource":{"primary":{"URL":"https:\/\/aapm.onlinelibrary.wiley.com\/doi\/10.1002\/mp.70665"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,9]]},"references-count":33,"journal-issue":{"issue":"9","published-print":{"date-parts":[[2026,9]]}},"alternative-id":["10.1002\/mp.70665"],"URL":"https:\/\/doi.org\/10.1002\/mp.70665","archive":["Portico"],"relation":{},"ISSN":["0094-2405","2473-4209"],"issn-type":[{"value":"0094-2405","type":"print"},{"value":"2473-4209","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,9]]},"assertion":[{"value":"2026-04-10","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-08-12","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-09-02","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70665"}}