Purpose/Significance To construct a dual-layer analytical framework to comprehensively understand disciplinary structures，identify emerging frontiers，and track topic evolution. Method/Process Medical informatics literatures from 2016 to 2025 are retrieved from PubMed，Scopus，and Web of Science databases. The topics are identified by BERTopic and classified into three evolution patterns: emerging，stable and declining. A retrieval-augmented generation (RAG) system is built based on ChromaDB，enabling micro-level validation and knowledge association mining through document-topic mapping. Result/Conclusion The evolution of medical informatics topics exhibits three major characteristics: research focus shift，deepening technological integration，and enhanced interdisciplinary convergence. The BERTopic-RAG framework provides a novel methodology for knowledge discovery.