How AI is used in FDA-authorized medical devices: a taxonomy across 1,016 authorizations

Abstract

Abstract The recent proliferation of AI-enabled medical devices and the growing emphasis on clinical translation creates a critical need to understand the evolving scope of these technologies. We reviewed 1,016 FDA authorizations of AI/ML-enabled medical devices to develop a taxonomy that captures key axes of variation in clinical and AI-related features. While quantitative image analysis remains the most common application, the relative proportion of such devices has declined recently due to an increase in devices designed for different data types and higher-risk use cases. Notably, over 100 devices leverage AI for data generation, including synthetic image creation, though we did not currently find evidence of LLM-based generative models. Nonetheless, the FDA’s scope of AI/ML appears broad, encompassing devices that explicitly reference traditional ML methods such as K-nearest neighbors. To accompany the analysis, we have created a website to facilitate exploration of our curated database and trends over time. Altogether, our taxonomy and findings clarify how AI is currently used in medical devices and provide a foundation for tracking future developments as clinical applications evolve.

Ähnliche Arbeiten