ARTICLE SUMMARY:
What are the current limitations of AI/ML technology, and what must industry stakeholders do to fully leverage the technology's continuous learning capabilities to increasingly improve clinical outcomes? Global regulatory and standards bodies are harmonizing to get us closer to tapping this infinite potential, and manufacturers will play a critical role in bringing this powerful promise to fruition. By Sharon Murphy and Yuan Fang, Regulatory Affairs, Hologic Inc.
The field of artificial intelligence machine learning (AI/ML) has come a long way, but it still has far to go before we can fully realize its promise in the use of medical products. Technically, AI/ML software can continuously learn as it goes, digesting each new data point in real time to inform and improve its own capabilities. However, most approved and pipeline medical products use “locked” versions of the software that do not learn on the go. So far, no reliable, proven way exists for AI to test and validate its own safety and efficacy with each iterative change.
