FDA to hold key public workshop on AI in Radiology
Presentations will feature pivotal conversations around software and device regulation
Bethesda — The Food and Drug Administration (FDA) has released a detailed itinerary of its upcoming public workshop to be held February 25–26 at the NIH Main Campus in Bethesda, Maryland. The two-day workshop promises to set the tone around regulatory standards for imaging software and hardware powered by artificial intelligence (AI). Because this is the first big regulatory discussion hosted by the FDA on AI in radiology to date, it’s probable that ideas and positions advocated by participants at this workshop will influence regulatory expectations for years to come . It is too late to register in person, but you might still be able to join via webinar by asking nicely — try emailing RadAIWorkshop2020@fda.hhs.gov.
Where AI has come to represent a nebulous collection of techniques that ultimately involve software-based automation informed by data, the sessions relate to one of two areas of developing products and regulatory oversight: image interpretation, and image acquisition.
Pivotal topics to be discussed in talks and panels
Among the types of radiology discussed, there is heavy emphasis on ultrasound technology, where emerging technologies offer both image interpretation, and real-time guidance during image acquisition. Other modalities like MRI and CT are to be included in discussion around image interpretation, although AI-optimized acquisition methods are relevant to image acquisition with these modalities as well, such as pulse sequence optimization, slice-selection guidance, dose minimization, and image reconstruction.
There are important regulatory questions at stake in this discussion that will ultimately impact public health:
- What constitutes sufficient validation for AI in radiology?
- What are the standards for post-market monitoring?
- What are appropriate training standards for people who use radiological AI systems?
- What are the appropriate methods to validate data security?
- What rights and compensation do patients have in the ongoing use of their imaging to enhance AI algorithms?
Participation emphasizes FDA and industry
Participation includes FDA regulators, NIH scientists, academic physicians, the ACR, some nonprofits, hospital systems, and industry, with the bulk of sessions coming from the FDA and industry.
Opening lectures include talks from these leading FDA representatives:
- Jennifer Segui, Lead Reviewer, Division of Radiological Health
- Alex Cadotte PhD, Division Team Lead, Software and Digital Health
- Bakul Patel, MSEE, MBA, Director, Division of Digital Health
- Robert Ochs, PhD, Deputy Director for Radiological Health, Office of In Vitro Diagnostics and Radiological Health (FDA)
- Jessica Lamb, PhD, Assistant Director (Acting), Mammography, Ultrasound, and Imaging Software Team, Division of Radiological Health (FDA)
- Nicholas Petrick, PhD, Deputy Director Division of Imaging, Diagnostics and Software Reliability, Office of Science and Engineering Laboratories
- Matthew Diamond, MD, PhD, Medical Officer, Digital Health, Center for Devices and Radiological Health
- Shahram Vaezy, PhD, Biomedical Engineer, Division of Radiological Health
Academic institutions represented at the two-day workshop include MGH, UC Irvine, University of Chicago, and Stanford. Hospital systems represented include Partners Healthcare and Intermountain Health.
Industry representation includes IBM Watson, GE Healthcare, Siemens, FUJIFILM, Imagen Technologies, Butterfly Networks, and others.
— Orbit Staff