First ever AI/ML action plan by FDA


The U.S. last week. The organization’s first Artificial Intelligence/Machine Learning (AI/ML)– Based Software as a Medical Device (SaMD) Action Plan was presented by the Food and Drug Administration. This proposal demonstrates a multi-pronged approach to deal with the monitoring of AI/ML-based medical software by the Department.

The Action Plan for Artificial Intelligence/Machine Learning (AI/ML)-Based Software as a Medical Device (SaMD) is a response to stakeholder feedback on the 2019 FDA regulatory framework for medical products based on AI and ML.

Besides, the FDA will conduct a public workshop on algorithm transparency and draw on other primary activities from its stakeholders and collaborators, such as algorithm predisposition assessment. Although a guide for propelling a regulatory system is suggested in the Action Plan, the organizational structure gives the impression of being further down the track.

The Medical Device Action Plan AI/ML-Based Program maps five steps that the FDA plans to take, including:

  • Further implementation of the proposed administrative structure, including the issuance of draft guidelines on a preordained change control plan (after some time for software learning).
  • Support the progress of good machine learning practices to test and strengthen ML algorithms.
  • Cultivating an approach based on patients, including accountability of the device to customers.
  • Creating methods to test and strengthen ML algorithms.
  • Propelling real-world pilots for performance tracking.

The FDA is preparing to publish it in 2021. Different areas of improvement would include improving the identification of the types of improvements necessary under the process, as well as information related to the focused evaluation, such as the accommodation period and the content of the submission.

The business will also seek to promote the implementation of good machine learning practices. The FDA noted that the turn of events and the adoption of AI/ML best practices is important not only for regulating product design but also for promoting supervision of these high-level devices. For AI and ML gadgets, which can learn and adjust over the long-term and consolidate algorithms that show a degree of haziness, the FDA noted that transparency is especially important.

In October 2020, the FDA held a Patient Experience Advisory Committee (PEAC) meeting to guarantee openness in AI and ML medical device software. Patients are also given feedback on what elements affect their faith in these developments.

The CDRH Digital Health Center of Excellence, launched in September 2020, is focused on strategically propelling science and evidence for digital health technology within the FDA’s administrative and oversight framework. The Center’s mission is to allow partners to promote medical care by fostering responsible and excellent innovation in digital health.

For AI/ML engineers, the Action Plan may seem modest in its 2021 destinations. The only clear obligation for 2021, for example, is to issue draft guidance on Predetermined Change Management Strategies, which is only a single aspect of the multi-pronged approach of the Agency distributed in its Discussion Paper. Engineers, however, should see this as an opportunity to draw on the FDA and influence the thinking of the department on key concepts that will eventually be incorporated into a comprehensive system.

 Follow and connect with us on Facebook, Linkedin & Twitter


Please enter your comment!
Please enter your name here