Mapping Open-Source Health AI: A Graph-Analytics Study Across Imaging, EHR, Genomics, and General Clinical Applications

Abstract

Open-source software has become a core driver of progress in health-focused artificial intelligence, yet the structure of its developer and project ecosystem remains poorly understood. This paper presents a graph analytics study of public GitHub repositories related to health AI, revealing how collaborative communities are organised across subfields. A dataset of repositories was assembled using a keyword-based search that combines health-related terms from different subdomains, such as imaging, genetics, and electronic health records, with AI-related terms. Automated filters retained active, substantive projects, while removing duplicates and off-topic entries. The resulting corpus formed a manageable slice of the health-AI landscape. The ecosystem was modelled as a repository-level collaboration network, where nodes represent repositories and edges encode shared contributors and parent–fork relationships, with edge weights reflecting collaboration strength. The study identified mature, densely connected communities alongside emerging or weakly connected areas, and outlines implications for researchers, funders, public code repository maintainers, and industry stakeholders seeking tools, collaborators, or underexplored niches.