{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"institution":[{"name":"medRxiv"}],"indexed":{"date-parts":[[2026,7,9]],"date-time":"2026-07-09T20:26:32Z","timestamp":1783628792757,"version":"3.55.0"},"posted":{"date-parts":[[2026,1,23]]},"group-title":"Cardiovascular Medicine","reference-count":41,"publisher":"openRxiv","license":[{"start":{"date-parts":[[2026,1,23]],"date-time":"2026-01-23T00:00:00Z","timestamp":1769126400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by-nc-nd\/4.0\/"}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"accepted":{"date-parts":[[2026,1,23]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                <jats:sec>\n                  <jats:title>Background<\/jats:title>\n                  <jats:p>Patients with repaired tetralogy of Fallot (rTOF) require lifelong surveillance with cardiovascular magnetic resonance (CMR) and cardiopulmonary exercise testing (CPET). However, results are frequently stored as unstructured free-text reports, hindering large-scale analysis and research.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Objectives<\/jats:title>\n                  <jats:p>This study aimed to develop and evaluate a privacy-preserving hybrid natural language processing pipeline combining regular expressions (regex) and an on-premises large language model (LLM) to accurately extract key CMR and CPET metrics from legacy free-text reports in patients with rTOF.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Methods<\/jats:title>\n                  <jats:p>We retrospectively analyzed 430 CMR and 262 CPET reports (2005\u20132023) from patients with rTOF. A two-stage hybrid pipeline was implemented: regex rules were applied first, followed by targeted prompting of an on-premises Llama-3.1-8B-Instruct LLM only when regex failed or returned ambiguous results. Performance was compared against regex-only and LLM-only approaches using coverage, precision, recall, and F1-score.<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Results<\/jats:title>\n                  <jats:p>In CMR reports, the hybrid pipeline achieved perfect coverage (1.00) and F1-score (1.00) versus 0.98 coverage and 0.93 F1-score with regex alone, while reducing computational cost by ~75% compared with LLM-only. In CPET reports, the hybrid approach improved F1-score from 0.74 (regex alone) to 0.98, with particularly large gains for semantically complex variables (e.g., peak VO \u25a1, test termination reason).<\/jats:p>\n                <\/jats:sec>\n                <jats:sec>\n                  <jats:title>Conclusions<\/jats:title>\n                  <jats:p>A hybrid regex\u2013on-premises LLM pipeline provides near-perfect, efficient, and HIPAA-compliant extraction of clinical metrics from unstructured cardiology reports, offering a scalable solution for retrospective research and quality improvement in congenital heart disease.<\/jats:p>\n                <\/jats:sec>","DOI":"10.64898\/2026.01.21.26344531","type":"posted-content","created":{"date-parts":[[2026,1,23]],"date-time":"2026-01-23T21:35:31Z","timestamp":1769204131000},"source":"Crossref","is-referenced-by-count":2,"title":["Hybrid rule-based and on-premises LLM pipeline for extracting CMR and CPET metrics from free-text reports in repaired tetralogy of Fallot"],"prefix":"10.64898","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-3055-7355","authenticated-orcid":false,"given":"Izzet T.","family":"Akbasli","sequence":"first","affiliation":[{"name":"Department of Heart, Vascular, and Thoracic, Division of Cardiology and Cardiovascular Medicine, Cleveland Clinic Children\u2019s. Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4603-0235","authenticated-orcid":false,"given":"Kristen L.","family":"Beck","sequence":"additional","affiliation":[{"name":"Digital Health, Algorithms and Applications, IBM Research, San Jose, California, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Wilson","family":"Liou","sequence":"additional","affiliation":[{"name":"Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sherry","family":"Du","sequence":"additional","affiliation":[{"name":"Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Gabriel","family":"Nemer","sequence":"additional","affiliation":[{"name":"Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5629-7788","authenticated-orcid":false,"given":"Orkun","family":"Baloglu","sequence":"additional","affiliation":[{"name":"Department of Integrated Hospital Care, Division of Critical Care. Children\u2019s Institute, Cleveland Clinic Children\u2019s. Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Samir Q.","family":"Latifi","sequence":"additional","affiliation":[{"name":"Department of Integrated Hospital Care, Division of Critical Care. Children\u2019s Institute, Cleveland Clinic Children\u2019s. Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI), Cleveland, Ohio, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Bradley S.","family":"Marino","sequence":"additional","affiliation":[{"name":"Department of Heart, Vascular, and Thoracic, Division of Cardiology and Cardiovascular Medicine, Cleveland Clinic Children\u2019s, Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI). Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5184-747X","authenticated-orcid":false,"given":"Samer","family":"Albahra","sequence":"additional","affiliation":[{"name":"Center for Diagnostics and Artificial Intelligence, Cleveland Clinic Foundation, Department of Pathology and Laboratory Medicine, Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9769-8801","authenticated-orcid":false,"given":"Animesh","family":"Tandon","sequence":"additional","affiliation":[{"name":"Department of Heart, Vascular, and Thoracic, Division of Cardiology and Cardiovascular Medicine, Cleveland Clinic Children\u2019s, Cleveland Clinic Children\u2019s Center for Artificial Intelligence (C4AI), Cleveland, OH, USA"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"54368","reference":[{"key":"2026012707050774000_2026.01.21.26344531v1.1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ahj.2005.03.058"},{"key":"2026012707050774000_2026.01.21.26344531v1.2","doi-asserted-by":"publisher","DOI":"10.1111\/echo.15486"},{"key":"2026012707050774000_2026.01.21.26344531v1.3","doi-asserted-by":"publisher","DOI":"10.1016\/j.jocmr.2024.101062"},{"key":"2026012707050774000_2026.01.21.26344531v1.4","doi-asserted-by":"publisher","DOI":"10.1161\/CIR.0000000000000602"},{"key":"2026012707050774000_2026.01.21.26344531v1.5","doi-asserted-by":"publisher","DOI":"10.1161\/CIR.0000000000001291"},{"key":"2026012707050774000_2026.01.21.26344531v1.6","first-page":"950","article-title":"Impact of Automated Transfer of Semi-Automated Segmentation and Structured Report Rule Requirements on Cardiac MRI Report Quality, Standardization, and Efficiency","volume":"2024","year":"2025","journal-title":"AMIA Annu Symp Proc"},{"key":"2026012707050774000_2026.01.21.26344531v1.7","doi-asserted-by":"publisher","DOI":"10.1148\/radiol.2531090138"},{"key":"2026012707050774000_2026.01.21.26344531v1.8","doi-asserted-by":"publisher","DOI":"10.1186\/s12911-022-02022-1"},{"key":"2026012707050774000_2026.01.21.26344531v1.9","unstructured":"Devlin J , Chang M-W , Lee K , Toutanova K (2019) BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding"},{"key":"2026012707050774000_2026.01.21.26344531v1.10","doi-asserted-by":"publisher","DOI":"10.1186\/s12911-025-02871-6"},{"key":"2026012707050774000_2026.01.21.26344531v1.11","doi-asserted-by":"publisher","DOI":"10.1016\/j.redii.2023.100027"},{"key":"2026012707050774000_2026.01.21.26344531v1.12","doi-asserted-by":"publisher","DOI":"10.2196\/59782"},{"key":"2026012707050774000_2026.01.21.26344531v1.13","doi-asserted-by":"publisher","DOI":"10.1056\/NEJMsr2214184"},{"key":"2026012707050774000_2026.01.21.26344531v1.14","doi-asserted-by":"publisher","DOI":"10.1148\/ryai.230364"},{"key":"2026012707050774000_2026.01.21.26344531v1.15","doi-asserted-by":"publisher","DOI":"10.1001\/jama.2023.9458"},{"key":"2026012707050774000_2026.01.21.26344531v1.16","unstructured":"Medicine NA of, Series TLHS , Whicher D , et al (2023) HEALTH CARE ARTIFICIAL INTELLIGENCE: LAW, REGULATION, AND POLICY. National Academies Press (US)"},{"key":"2026012707050774000_2026.01.21.26344531v1.17","unstructured":"Elm E von , Altman DG , Egger M , et al The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies"},{"key":"2026012707050774000_2026.01.21.26344531v1.18","unstructured":"Wang X , Liu G , Zhu B , et al (2025) Pre-trained Language Models and Few-shot Learning for Medical Entity Extraction"},{"key":"2026012707050774000_2026.01.21.26344531v1.19","doi-asserted-by":"publisher","DOI":"10.4274\/dir.2024.242702"},{"key":"2026012707050774000_2026.01.21.26344531v1.20","doi-asserted-by":"publisher","DOI":"10.1161\/CIR.0b013e3181e52e69"},{"key":"2026012707050774000_2026.01.21.26344531v1.21","doi-asserted-by":"publisher","DOI":"10.1067\/mje.2002.121536"},{"key":"2026012707050774000_2026.01.21.26344531v1.22","unstructured":"Santana De Almeida S , Silva Fontes R , Pareja Credidio Freire Alves L , et al Frontiers | Artificial Intelligence in Healthcare Text Processing: A Review Applied to Named Entity Recognition"},{"key":"2026012707050774000_2026.01.21.26344531v1.23","doi-asserted-by":"publisher","DOI":"10.1055\/s-0044-1778693"},{"key":"2026012707050774000_2026.01.21.26344531v1.24","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-025-01686-z"},{"key":"2026012707050774000_2026.01.21.26344531v1.25","doi-asserted-by":"publisher","DOI":"10.2196\/64226"},{"key":"2026012707050774000_2026.01.21.26344531v1.26","unstructured":"Islam MA , Jonnala DV , Rekhi R , et al (2025) Evaluating the Energy-Efficiency of the Code Generated by LLMs"},{"key":"2026012707050774000_2026.01.21.26344531v1.27","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-025-01670-7"},{"key":"2026012707050774000_2026.01.21.26344531v1.28","doi-asserted-by":"publisher","DOI":"10.1016\/S2589-7500(23)00083-3"},{"key":"2026012707050774000_2026.01.21.26344531v1.29","doi-asserted-by":"publisher","DOI":"10.1016\/j.neucom.2024.128096"},{"key":"2026012707050774000_2026.01.21.26344531v1.30","doi-asserted-by":"publisher","DOI":"10.1007\/s00330-023-09901-9"},{"key":"2026012707050774000_2026.01.21.26344531v1.31","doi-asserted-by":"publisher","DOI":"10.1093\/jamia\/ocae069"},{"key":"2026012707050774000_2026.01.21.26344531v1.32","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-024-80165-z"},{"key":"2026012707050774000_2026.01.21.26344531v1.33","doi-asserted-by":"publisher","DOI":"10.1016\/j.patter.2024.101006"},{"key":"2026012707050774000_2026.01.21.26344531v1.34","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-024-01126-4"},{"key":"2026012707050774000_2026.01.21.26344531v1.35","doi-asserted-by":"crossref","unstructured":"Jung J , Jeong H , Huh E-N (2024) Federated Learning and RAG Integration: A Scalable Approach for Medical Large Language Models","DOI":"10.1109\/ICAIIC64266.2025.10920885"},{"key":"2026012707050774000_2026.01.21.26344531v1.36","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-025-01661-8"},{"key":"2026012707050774000_2026.01.21.26344531v1.37","doi-asserted-by":"publisher","DOI":"10.1093\/ehjdh\/ztac047"},{"key":"2026012707050774000_2026.01.21.26344531v1.38","first-page":"4125","article-title":"Translational NLP: A New Paradigm and General Principles for Natural Language Processing Research","volume":"2021","year":"2021","journal-title":"Proc Conf Assoc Comput Linguist North Am Chapter Meet"},{"key":"2026012707050774000_2026.01.21.26344531v1.39","doi-asserted-by":"publisher","DOI":"10.2196\/44977"},{"key":"2026012707050774000_2026.01.21.26344531v1.40","unstructured":"Yee JSG , Ng PC , Wang Z , et al (2024) On-Device LLMs for SMEs: Challenges and Opportunities"},{"key":"2026012707050774000_2026.01.21.26344531v1.41","doi-asserted-by":"publisher","DOI":"10.3390\/info16020087"}],"container-title":[],"original-title":[],"link":[{"URL":"https:\/\/syndication.highwire.org\/content\/doi\/10.64898\/2026.01.21.26344531","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,1,27]],"date-time":"2026-01-27T15:23:38Z","timestamp":1769527418000},"score":1,"resource":{"primary":{"URL":"http:\/\/medrxiv.org\/lookup\/doi\/10.64898\/2026.01.21.26344531"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,23]]},"references-count":41,"URL":"https:\/\/doi.org\/10.64898\/2026.01.21.26344531","relation":{},"subject":[],"published":{"date-parts":[[2026,1,23]]},"subtype":"preprint"}}