Medi-Sum: Smart Medical Document Synthesis System

Abstract

Medical prescriptions and clinical documents are often difficult for patients to understand due to complex medical terminology, handwritten content, and language barriers. This paper presents Medi-Sum, an AI-powered Android-based medical document synthesis system that converts unstructured prescriptions into structured and patient-friendly health information. The proposed system employs a hybrid Optical Character Recognition (OCR) framework combining Google Gemini Vision API, Microsoft TrOCR, and PyTesseract to accurately extract text from both printed and handwritten prescriptions. Extracted information is processed using a Large Language Model to perform medical entity extraction and generate separate summaries for healthcare professionals and patients. A RetrievalAugmented Generation (RAG)-based chatbot enables users to ask health-related questions using their prescription history while supporting multiple Indian regional languages. The system also provides medication reminders, diagnostic test alerts, health timeline management, vitals tracking, and generic medicine recommendations within a privacy-focused architecture. Experimental evaluation on 500 anonymized clinical documents demonstrated high extraction accuracy, effective summarization quality, and reliable chatbot performance, indicating the suitability of the proposed approach for improving patient understanding, healthcare accessibility, and continuity of care in multilingual and resource-constrained environments. Keywords— Prescription Digitization, Optical Character Recognition (OCR), Large Language Models (LLMs), Named Entity Recognition (NER), Retrieval-Augmented Generation (RAG), Medical Document Summarization, Multilingual Healthcare, Medical Chatbot, Patient Health Records, Android Healthcare Application, Clinical Decision Support, Healthcare Informatics.