Agency Takes Steps to Promote Scientifically Sound Development of AI-Based Medical Devices

The FDA is stepping up its efforts to regulate the development of medical devices that use artificial intelligence and machine learning (AI/ML) technologies. Here’s a briefing on a pair of significant initiatives that the agency unveiled in the past six weeks.

The AI/ML Action Plan

On Sept. 22, the agency published an Action Plan setting forth five actions it intends to take to advance sound, safe and effective development of AI/ML-based medical devices:

1. Issue draft guidance for modification of algorithms in AI/ML software as a medical device (SaMD) over time;
2. Promote Good Machine Learning Practice (GMLP), i.e., AI/ML best practices for data management, feature extraction, training, interpretability and evaluation that are akin to good software engineering practices;
3. Hold a public workshop on how device labeling supports transparency to users and enhances trust in AI/ML-based devices;
4. Support regulatory science efforts to develop methodology for evaluating and improving machine learning algorithms, including via identification and elimination of bias; and
5. Support the piloting of real-world performance (RWP) monitoring for SaMD.

The AI/ML Best Practices Guidelines

On Oct. 27, the FDA and its counterpart agencies in Canada and the United Kingdom issued joint guidelines for companies developing The FDA, Health Canada and U.K. Medicines and Healthcare Products Regulatory Agency outline 10 “guiding principles” for companies that develop AI/ML-based medical devices:

1. Leverage multi-disciplinary expertise throughout the total product life cycle;
2. Implement Good Software Engineering and Security Practices;
3. Ensure that relevant characteristics of the intended patient population (e.g., regarding age, gender, sex, race, and ethnicity), use and measurement inputs are sufficiently represented in a sample of adequate size in the clinical study and training and test datasets, so that results can be reasonably generalized to the population of interest;
4. Ensure that training and test datasets are appropriately independent of one another;
5. Select referenced datasets upon best available methods;
6. Tailor model design to the available data and ensure that it reflects the device’s intended use;
7. Account for human factors in evaluating performance of models that include a “human in the loop”;
8. Ensure that testing of device performance is carried out in clinically relevant conditions;
9. Provide device users clear, essential information; and
10. Monitor models that are deployed in real world condition with a focus on maintaining or improving safety and performance.

As the AI/ML medical device field evolves, so too must [Good Machine Learning Practice] best practice and consensus standards, the new guidance notes.

Here are some of the key new FDA EUAs and clearances announced in October 2021:

New FDA Emergency Use Authorizations (EUAs) & Approvals

New CE Marks & Global Certifications

Notable European CE certifications announced during the period:

NEW CE MARKINGS IN EUROPE

Other international clearances announced during the period: