Med Intel: An AI-Based Healthcare Pharmacy Management System with Intelligent Prescription Recognition and Drug Safety Verification

DOI

https://www.doi.org/10.59256/ijsreat.20260606001

Published In

Volume 6, Issue 6 (2026)

Published Online

November-December 2026

Pages

01-05

Article Type

Original Article

Check for updates

The manual processing of handwritten prescriptions and the lack of automated drug safety verification in pharmacy management systems pose significant risks to patient safety and operational efficiency. This paper presents MedIntel, an AI-based healthcare pharmacy management system that integrates intelligent prescription recognition and comprehensive drug interaction detection. The prescription recognition module employs a dual-engine optical character recognition (OCR) pipeline combining EasyOCR with Tesseract as a fallback, enhanced by four image preprocessing strategies: grayscale conversion, contrast-limited adaptive histogram equalization (CLAHE), median filtering denoising, and Otsu binary thresholding. Extracted text undergoes a multi-stage natural language processing pipeline that splits recognized text into individual word candidates, filters noise using stop-word elimination and alphabetic character ratio thresholds, and matches each candidate against the medicine database using prefix-based SQL ILIKE queries augmented by RapidFuzz token-based fuzzy matching. The drug interaction detection system performs real-time pairwise checking of all cart items against a pre-populated DrugInteraction database using bidirectional ILIKE fuzzy matching, presenting warnings categorized by four severity levels: critical, severe, moderate, and mild, each accompanied by clinical descriptions and actionable recommendations. The system is developed using Python with Flask framework on the backend and Bootstrap 5.3 on the frontend, with SQLite for persistent storage. Evaluation of the medicine name recognition module achieves a token-level accuracy of 99.19% with an entity-level F1-score of 0.971 across 140 annotated prescription sentences, and the drug interaction database covers 15 clinically validated interaction pairs across four severity categories. The system demonstrates the feasibility of deploying lightweight, database-driven AI approaches for enhancing pharmacy safety in resource-constrained healthcare environments.