DEVELOPMENT OF A PROGRAM FOR ANALYZING MEDICAL LABORATORY RESULTS USING ARTIFICIAL INTELLIGENCE MODELS
DOI:
https://doi.org/10.5281/zenodo.19606581Keywords:
EasyOCR, JSON, LLM, OLM, Ollama, laboratory diagnostics, artificial intelligence, NLP, CDSS, RAG, medical data, clinical decision supportAbstract
This paper presents the concept and architecture of an intelligent system for automatic processing of medical
laboratory test results. The system integrates three key components: EasyOCR-based optical character recognition
for extracting data from scanned forms, structured JSON storage for normalized results, and a local language model
(OLM) via the Ollama framework for generating personalized medical recommendations. The application of Retrieval-
Augmented Generation improves the clinical accuracy of recommendations to 4.6 out of 5.0 points. The achieved OCR
accuracy of 95.5% for numerical fields meets ISO 15189 requirements. Total processing time per form does not exceed
17 seconds on CPU. The key advantage is a fully local architecture ensuring compliance with GDPR, HIPAA, and Federal
Law No. 152-FZ on personal data.
References
WHO. Laboratory services in support of primary health care. — Geneva: World Health Organization, 2022.
Baek Y. et al. Character region awareness for text detection // Proceedings of CVPR. — 2019. — P. 9365–9374.
HL7 International. FHIR R5 Specification: Observation Resource. — 2023. — URL: https://hl7.org/fhir/observation.
html
Singhal K. et al. Large Language Models Encode Clinical Knowledge // Nature. — 2023. — Vol. 620. — P. 172–180.
Lewis P. et al. Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks // Advances in NeurIPS. — 2020.
— Vol. 33.
CALIPER. Canadian Laboratory Initiative on Pediatric Reference Intervals. — University of Toronto, 2022
ISO 15189:2022. Medical laboratories — Requirements for quality and competence. — Geneva: ISO, 2022.
Vaswani A. et al. Attention Is All You Need // Advances in NeurIPS. — 2017. — Vol. 30.
JaidedAI. EasyOCR: Ready-to-use OCR with 80+ supported languages. — GitHub, 2024.
Federal Law No. 152-FZ of 27.07.2006 «On Personal Data». — Moscow, 2006.
Qwen Team. Qwen2.5 Technical Report. — Alibaba Cloud, 2024.
HIMSS. Clinical Decision Support Toolkit. — Chicago: HIMSS, 2023.
