ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING IN MEDICAL DEVICES: FDA REGULATORY PATHWAYS, TRENDS, AND THE CASE OF NEMOSCAN 510(K) CLEARANCE

Abstract

With the introduction of artificial intelligence (AI) and machine learning (ML) in medical device software, in particular software as a medical device (SaMD), healthcare continues to be reshaped. This review provides a historical overview of AI and the foundational principles of ML as well as the regulatory framework it currently undergoes to achieve all the medical devices enabled by AI/ML. It outlines 510(k), De Novo, and PMA regulatory pathways and recent efforts made to muster Good Machine Learning Practice (GMLP) on a global scale. The study also explores the tendencies of FDA approvals, especially in terms of device classification and manufacturer impact, as well as regulatory openness. Described in this case study is how the NemoScan (K232698) dental planning software has been brought to light as an example of practical FDA 510(k) clearance of the substantial equivalence. To ensure that upcoming technological changes and further use of AI/ML-based medical hardware turn out to be not only safe but also effective, the review culminates by mentioning disaster-based regulation and empirical data, as well as interdisciplinary collaboration.

KEYWORDS:

510(k) Clearance, Artificial Intelligence, Machine Learning, Medical Devices, Clinical Trials