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

The researchers reviewed 1,016 FDA-authorized AI/ML-enabled medical devices (from 1995 to 2024) and developed a detailed taxonomy to classify how AI is currently being used in clinical devices.

The taxonomy has three core factors:

• The first factor is the core clinical function of the device, describing the general medical role or purpose it serves in patient care.
• The second factor is the AI function, capturing how the device uses AI to assist with the clinical function.
• The third factor is the data type used as input to perform the AI function.

For data type, 84.4% of devices rely on image data, predominantly for radiology.

In terms of clinical function, the majority of current devices fall under Assessment (84.1%), such as diagnosis or monitoring.

For AI function, the majority (85.6%) use AI for data analysis, particularly for quantification/feature localization and triage.

The taxonomy clarifies current AI usage in medical devices and provides a foundation for tracking developments as applications evolve.