A Conversational Agent for Rare Eye Disease Knowledge Support in Local Italian Healthcare Contexts: The ELENA Framework

Accessing global medical knowledge on rare ophthalmic diseases remains a challenge for small healthcare centers, as relevant literature and case reports are scattered across multiple languages and formats. To bridge this gap, we introduce ELENA (Eye-disease Language-model for Evidence-based Navigation and Assistance), a retrieval-augmented chatbot designed to provide multilingual access to specialized medical knowledge in ophthalmic rare diseases. ELENA combines a quantized Large Language model with a lightweight retrieval-augmented generation pipeline, where all source documents are pre-translated into English to ensure consistent embedding and retrieval. Italian user queries are automatically translated, processed through the pipeline, and the generated responses are translated back into Italian, yielding fluent and contextually grounded answers. Experiments on a curated corpus of ophthalmic rare-disease documents confirm the technical feasibility of deploying such systems even on modest hardware, while a graphical interface ensures accessibility for non-technical clinical staff. ELENA thus demonstrates how an efficient, privacy-preserving, and humanfriendly AI assistant can democratize access to international medical knowledge within local healthcare environments.