Artificial Intelligence in Valvular Heart Disease: A Bibliometric and Knowledge-Mapping Analysis Revealing a Three-Phase Evolution Toward Integrated Clinical Intelligence

The application of artificial intelligence (AI) in valvular heart disease (VHD) has advanced significantly in recent years. However, a systematic evaluation of this field's historical development, knowledge structure, and global research patterns has yet to be conducted. This paper aims to outline the global evolution of AI research in VHD utilizing bibliometric and knowledge-mapping methods.

Publications related to AI and VHD from 1991 to 2025 were retrieved from the Web of Science Core Collection database to investigate research progress. CiteSpace software was employed for bibliometric and knowledge mapping analyses, assessing indicators such as publication trends, keyword networks, collaboration patterns, co-citation relationships, and emerging topics.

A total of 3291 pieces of literature were identified. Research productivity remained low until the year before 2004, followed by steady growth and a sharp increase after 2015, culminating in 461 articles by the end of 2025. Network analysis revealed three sequential stages: image-assisted applications, prediction modeling, and recent advancements in deep-learning-based integrated systems. The United States accounted for the largest research volume, while England and Germany served as core nodes in the international cooperation network. Co-citation analysis identified significant clinical guidelines and pioneering interventional studies as key intellectual pillars in this domain.

Over the past 35 years, significant progress has been achieved, transitioning from computer systems used solely for image analysis to more integrated clinical applications. Currently, deep learning techniques are increasingly prevalent, accompanied by a focus on multimodal data integration and a stronger linkage between computational methods and clinical decision-making.