In April, the FDA issued a discussion paper and request for feedback regarding this proposed framework, which would regulate the use of medical device algorithms that use machine learning to continually update as they are fed more data. Currently, the FDA only authorizes the use of AI products that operate on a fixed algorithm.
After “applauding” the FDA’s drafted framework, the AMIA proposed four major modifications:
- A stronger emphasis on the stark difference between the two types of algorithms.
- Acknowledgement of the risk that machine learning algorithms can “learn” from weak or biased data and are not always able to explain their decisions.
- A warning about the possibility that algorithms can be exposed to cybersecurity risks with unregulated updates.
- Provisions to ensure algorithms are developed to be as unbiased as possible.
More articles about AI:
AI tool helps clinicians, parents distinguish between infants’ cries
Smartphone app ‘listens’ to coughs to diagnose respiratory disorders
Researchers training AI mobile app to detect early signs of oral cancer
