Many FDA-approved AI medical devices are not trained on real patient data

Researchers found that nearly half of FDA-approved AI medical devices have not been trained on real patient data, raising concerns about their clinical effectiveness. A cross-institutional study led by UNC and Duke University analyzed over 500 medical AI devices, finding a need for increased transparency and higher standards for clinical validation.

• AI has broad applications in healthcare.
• FDA approvals increased since 2016.
• 43% lack clinical validation data.
• Researchers propose new standards.
• Study published in Nature Medicine.