{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,10,21]],"date-time":"2026-10-21T18:43:05Z","timestamp":1792608185387,"version":"4.5.0"},"reference-count":27,"publisher":"JMIR Publications Inc.","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["JMIR Med Inform"],"abstract":"Abstract<\/jats:title>\n \n Background<\/jats:title>\n The development of structured optical character recognition (OCR) systems for Korean medical documents is hindered by strict privacy regulations that limit access to real clinical data. Existing synthetic document frameworks are primarily designed for generic layouts and Western languages, and fail to capture the complex formatting, bilingual terminology, and hierarchical field semantics required for clinical OCR and document-understanding tasks in hospital environments.<\/jats:p>\n <\/jats:sec>\n \n Objective<\/jats:title>\n This study aimed to develop and validate a privacy-preserving, template-driven pipeline for generating realistic synthetic Korean medical documents with structured XML annotations, enabling OCR and document-understanding model training in settings where access to real clinical data is restricted.<\/jats:p>\n <\/jats:sec>\n \n Methods<\/jats:title>\n We designed MedDocSynth, a 7-stage synthetic document generation pipeline that reconstructs authentic medical document layouts from publicly available references, incorporates clinician-guided template selection, and uses a structured placeholder system encoding 17 medical data types. Human-curated templates were populated with exemplar content and tagged using a delimiter-based scheme, after which GPT-4o-mini generated diverse synthetic values for each field. The pipeline produced 1000 fully synthetic DOCX documents with corresponding XML annotations encoding hierarchical structure, title-answer relationships, and datatype metadata. We evaluated text fidelity, lexical diversity (type-token ratio, measure of textual lexical diversity [MTLD], n-grams, and Pareto), and compared the system with existing synthetic document pipelines. A Qwen2.5-VL model was fine-tuned exclusively on the synthetic dataset and evaluated on 50 real-world Korean clinical documents to assess OCR accuracy, information extraction, and table reconstruction.<\/jats:p>\n <\/jats:sec>\n \n Results<\/jats:title>\n The pipeline achieved 100% placeholder coverage, 99.68% format correctness, and 91.56% semantic accuracy. Lexical diversity analysis demonstrated high variability in free-text fields (MTLD\u2248456) and expected repetitiveness in structured datatypes. Compared with existing pipelines (SynthText, SynthTiger, and DocSynth), the proposed system uniquely supports complex clinical forms, bilingual content, and structured XML annotations. When trained solely on the synthetic data, Qwen2.5-VL achieved the highest real-world performance, with a character accuracy of 86.01%, word accuracy of 78.25%, field accuracy of 67.18%, datatype accuracy of 64.14%, and table reconstruction tree-edit-distance similarity score of 71.09%. Models without fine-tuning performed substantially worse, including GPT-5-mini.<\/jats:p>\n <\/jats:sec>\n \n Conclusions<\/jats:title>\n We present the first synthetic data generation pipeline specifically tailored to Korean medical documents, enabling the creation of realistic, richly annotated datasets for structured OCR without exposing patient information. The synthetic data significantly improves OCR performance on real clinical documents, demonstrating the pipeline\u2019s potential utility for hospital deployment and its potential extensibility to other languages and health care systems.<\/jats:p>\n <\/jats:sec>","DOI":"10.2196\/91313","type":"journal-article","created":{"date-parts":[[2026,8,13]],"date-time":"2026-08-13T01:10:08Z","timestamp":1786583408000},"page":"e91313-e91313","source":"Crossref","is-referenced-by-count":0,"title":["Synthetic Medical Document Generation Pipeline (MedDocSynth) for Optical Character Recognition: Development and Validation Study on Korean Clinical Documents"],"prefix":"10.2196","volume":"14","author":[{"ORCID":"https:\/\/orcid.org\/0009-0003-0940-3834","authenticated-orcid":false,"given":"Mikel","family":"Lainsa","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0007-6802-944X","authenticated-orcid":false,"given":"Javier","family":"Aguirre","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0000-1048-8386","authenticated-orcid":false,"given":"Moon","family":"Hyeyoon","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0008-6118-3687","authenticated-orcid":false,"given":"Won","family":"Yejun","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0003-9672-2250","authenticated-orcid":false,"given":"Woojin","family":"Jang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1058-990X","authenticated-orcid":false,"given":"Jong","family":"Eun Park","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-2778-2992","authenticated-orcid":false,"given":"Won","family":"Chul Cha","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"1010","published-online":{"date-parts":[[2026,10,21]]},"reference":[{"issue":"15","key":"R1","doi-asserted-by":"publisher","first-page":"1486","DOI":"10.1056\/NEJMlim035027","article-title":"HIPAA regulations - a new era of medical-record privacy?","volume":"348","author":"Annas","journal-title":"N Engl J Med"},{"issue":"1","key":"R2","doi-asserted-by":"publisher","first-page":"447","DOI":"10.1007\/s41870-023-01610-2","article-title":"OCR-MRD: performance analysis of different optical character recognition engines for medical report digitization","volume":"16","author":"Batra","journal-title":"Int j inf tecnol"},{"issue":"1","key":"R3","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-019-0190-1","article-title":"Challenges of developing a digital scribe to reduce clinical documentation burden","volume":"2","author":"Quiroz","journal-title":"NPJ Digit Med"},{"issue":"3","key":"R4","doi-asserted-by":"publisher","first-page":"446","DOI":"10.1055\/s-0044-1787007","article-title":"Toward alleviating clinician documentation burden: a scoping review of burden reduction efforts","volume":"15","author":"Sloss","journal-title":"Appl Clin Inform"},{"issue":"1","key":"R5","doi-asserted-by":"publisher","first-page":"129","DOI":"10.1146\/annurev-statistics-040720-031848","article-title":"Synthetic data","volume":"8","author":"Raghunathan","journal-title":"Annu Rev Stat Appl"},{"key":"R6","doi-asserted-by":"crossref","unstructured":"Gupta A Vedaldi A Zisserman A . Synthetic data for text localisation in natural images. Presented at: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR); Jun 27-30, 2016; Las Vegas, NV, USA. [doi: 10.1109\/CVPR.2016.254]","DOI":"10.1109\/CVPR.2016.254"},{"key":"R7","doi-asserted-by":"crossref","unstructured":"Yim M Kim Y Cho HC Park S . SynthTIGER: synthetic text image generator towards better text recognition models. Presented at: ICDAR 2021: 16th International Conference; Sep 5-10, 2021; Lausanne, Switzerland. [doi: 10.1007\/978-3-030-86337-1_8]","DOI":"10.1007\/978-3-030-86337-1_8"},{"key":"R8","doi-asserted-by":"crossref","unstructured":"Biswas S Riba P Llad\u00f3s J Pal U . DocSynth: a layout guided approach for controllable document image synthesis. Presented at: ICDAR 2021: 16th International Conference; Sep 5-10, 2021; Lausanne, Switzerland. [doi: 10.1007\/978-3-030-86334-0_36]","DOI":"10.1007\/978-3-030-86334-0_36"},{"key":"R9","unstructured":"OCR dataset. AI Hub. URL: https:\/\/www.aihub.or.kr [Accessed 10-10-2026]"},{"key":"R10","unstructured":"NAVER CLOVA AI. CLOVA OCR Service and Sample Data; n.d NAVER Corp, official demo site for CLOVA OCR. URL: https:\/\/api.ncloud-docs.com\/docs\/en\/ai-application-service-ocr. https:\/\/demo.ocr.clova.ai"},{"key":"R11","unstructured":"Lee S Kim JH Jung K Kim H . KAIST scene text database for automatic text location in scene images. Presented at: 2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPRW). 1-7; 2010.[doi: 10.1109\/CVPRW.2010.5543854]"},{"key":"R12","doi-asserted-by":"crossref","unstructured":"Cheng F Yada S Tanaka R Aramaki E Kurohashi S . JaMIE: a pipeline Japanese medical information extraction system with novel relation annotation. Presented at: Thirteenth Language Resources and Evaluation Conference; Jun 20-25, 2022; Marseille, France. [doi: 10.63317\/4t4c73deix4s]","DOI":"10.63317\/4t4c73deix4s"},{"key":"R13","unstructured":"Alshaikhdeeb B Hemedan AA Ghosh S Balaur I Satagopam V . Generation of synthetic clinical text: a systematic review. arXiv. Preprint posted online on Jul 24, 2025. [doi: 10.48550\/arXiv.2507.18451]"},{"key":"R14","doi-asserted-by":"publisher","DOI":"10.2196\/48443","article-title":"Unlocking the secrets behind advanced artificial intelligence language models in deidentifying Chinese-English mixed clinical text: development and validation study","volume":"26","author":"Lee","journal-title":"J Med Internet Res"},{"key":"R15","doi-asserted-by":"crossref","unstructured":"Zhong X Tang J Jimeno Yepes A . PubLayNet: largest dataset ever for document layout analysis. Presented at: 2019 International Conference on Document Analysis and Recognition (ICDAR); Sep 20-25, 2019; Sydney, Australia. [doi: 10.1109\/ICDAR.2019.00166]","DOI":"10.1109\/ICDAR.2019.00166"},{"key":"R16","doi-asserted-by":"crossref","unstructured":"Li M Xu Y Cui L et al. DocBank: a benchmark dataset for document layout analysis. Presented at: Proceedings of the 28th International Conference on Computational Linguistics; Dec 8-13, 2020; Barcelona, Spain (Online). [doi: 10.18653\/v1\/2020.coling-main.82]","DOI":"10.18653\/v1\/2020.coling-main.82"},{"key":"R17","doi-asserted-by":"crossref","unstructured":"Pfitzmann B Auer C Dolfi M Nassar AS Staar P . DocLayNet: a large human-annotated dataset for document-layout segmentation. Presented at: KDD \u201922: Proceedings of the 28th ACM SIGKDD Conference on Knowledge Discovery and Data Mining; Aug 14-18, 2022; Washington DC USA. [doi: 10.1145\/3534678.3539043]","DOI":"10.1145\/3534678.3539043"},{"issue":"1","key":"R18","doi-asserted-by":"publisher","DOI":"10.2196\/76773","article-title":"Toward cross-hospital deployment of natural language processing systems: model development and validation of fine-tuned large language models for disease name recognition in Japanese","volume":"13","author":"Shimizu","journal-title":"JMIR Med Inform"},{"issue":"1","key":"R19","doi-asserted-by":"publisher","DOI":"10.1038\/s41598-026-36904-5","article-title":"Bridging the performance gap: systematic optimization of local LLMs for Japanese medical PHI extraction","volume":"16","author":"Wada","journal-title":"Sci Rep"},{"key":"R20","doi-asserted-by":"crossref","unstructured":"Kumichev G Blinov P Kuzkina Y et al. MedSyn: LLM-based synthetic medical text generation framework. Presented at: Joint European Conference on Machine Learning and Knowledge Discovery in Databases; Sep 9-13, 2024; Vilnius, Lithuania. [doi: 10.1007\/978-3-031-70381-2_14]","DOI":"10.1007\/978-3-031-70381-2_14"},{"key":"R21","doi-asserted-by":"crossref","unstructured":"Ding C Liu X Tang W et al. SynthDoc: bilingual documents synthesis for visual document understanding. Presented at: LGM3A \u201924: Proceedings of the 2nd Workshop on Large Generative Models Meet Multimodal Applications; Oct 28 to Nov 1, 2024; Melbourne VIC Australia. [doi: 10.1145\/3688866.3689125]","DOI":"10.1145\/3688866.3689125"},{"key":"R22","doi-asserted-by":"crossref","unstructured":"Kim G Hong T Yim M et al. OCR-free document understanding transformer. Presented at: European Conference on Computer Vision; Oct 23-27, 2022; Tel Aviv, Israel. [doi: 10.1007\/978-3-031-19815-1_29]","DOI":"10.1007\/978-3-031-19815-1_29"},{"key":"R23","doi-asserted-by":"crossref","unstructured":"Nassar A Omenetti M Lysak M et al. SmolDocling: an ultra-compact vision-language model for end-to-end multi-modal document conversion. Presented at: 2025 IEEE\/CVF International Conference on Computer Vision (ICCV); Oct 21-23, 2025; Honolulu, HI, USA. [doi: 10.1109\/ICCV51701.2025.02040]","DOI":"10.1109\/ICCV51701.2025.02040"},{"key":"R24","doi-asserted-by":"publisher","DOI":"10.1016\/j.compbiomed.2025.109834","article-title":"Generative AI for synthetic data across multiple medical modalities: a systematic review of recent developments and challenges","volume":"189","author":"Ibrahim","journal-title":"Comput Biol Med"},{"issue":"3","key":"R25","doi-asserted-by":"publisher","first-page":"223","DOI":"10.1080\/09296174.2014.911506","article-title":"Can type-token ratio be used to show morphological complexity of languages?","volume":"21","author":"Kettunen","journal-title":"J Quant Linguist"},{"issue":"1","key":"R26","doi-asserted-by":"publisher","DOI":"10.1038\/s41746-020-0267-x","article-title":"Generation and evaluation of artificial mental health records for natural language processing","volume":"3","author":"Ive","journal-title":"NPJ Digit Med"},{"key":"R27","doi-asserted-by":"publisher","DOI":"10.1016\/j.csbj.2025.05.026","article-title":"Synthetic data in medicine: legal and ethical considerations for patient profiling","volume":"28","author":"Nisevic","journal-title":"Comput Struct Biotechnol J"}],"container-title":["JMIR Medical Informatics"],"original-title":[],"language":"en","deposited":{"date-parts":[[2026,10,21]],"date-time":"2026-10-21T18:12:32Z","timestamp":1792606352000},"score":1,"resource":{"primary":{"URL":"https:\/\/medinform.jmir.org\/2026\/1\/e91313"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10,21]]},"references-count":27,"URL":"https:\/\/doi.org\/10.2196\/91313","relation":{},"ISSN":["2291-9694"],"issn-type":[{"value":"2291-9694","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,10,21]]},"article-number":"v14i15e91313"}}