Abstract

MediScan: AI Health Report Analyzer & Advisor is a novel healthcare technology solution designed to extract, analyze, and interpret medical health reports in a structured and intelligent manner. The system leverages Optical Character Recognition (OCR) to digitize scanned reports and domain specific Natural Language Processing (NLP) models to identify clinical entities such as test names, values, reference ranges, and doctor remarks. By integrating a curated medical knowledge base with a hybrid rule-based and inference engine, MediScan automatically flags abnormalities, maps them to possible medical conditions, and generates personalized diet recommendations alongside non-prescriptive medicine references. The platform is built with strong emphasis on transparency, security, and patient privacy. Each abnormality detection and condition inference is auditable, referencing both extracted ranges and knowledge base guidelines. Medicine and diet recommendation engines are strictly advisory in nature, providing context-aware suggestions while strongly emphasizing consultation with qualified healthcare professionals. The technical stack incorporates Python (FastAPI backend), Tesseract OCR, Bio/Clinical BERT for NLP, and PostgreSQL for the medical knowledge base, with a React-based front-end interface. Ethical safeguards such as secure storage, ephemeral uploads, encryption, and strict disclaimers are integrated into the system design. MediScan addresses the critical need for accessible, AI-driven health insights by bridging raw medical data and actionable understanding. It aims to empower patients with quick, reliable interpretations while supporting clinicians with structured evidence, thus enhancing healthcare decision-making.