The clinician-AI interface: intended use and explainability in FDA-cleared AI devices for medical image interpretation

What the paper explores

The authors curated information on FDA-cleared AI devices for medical image interpretation, focusing on intended clinical use, the outputs presented to clinicians and the explanations provided. The resulting landscape shows variation in workflow roles, including triage, and in the information available to support interpretation.

How to interpret the evidence

This is a description of a historical device landscape. The presence of an explanation feature does not demonstrate that the explanation improves decisions or patient outcomes.

Questions for MedTech teams

Evaluate how users act on the output. Workflow integration, reliance, error handling and interpretability should be considered alongside model discrimination.