October 27, 2017

ACR Data Science Institute Meets With FDA on AI Regulatory Issues

ACR Data Science InstituteTM (DSI) efforts to advance development of safe and effective artificial intelligence (AI) solutions for medical imaging and radiological sciences are picking up speed.
ACR Data Science Institute™ (DSI) efforts to advance development of safe and effective artificial intelligence (AI) solutions for medical imaging and radiological sciences are picking up speed.

DSI leaders recently met with representatives from the Food and Drug Administration (FDA) about the regulatory review process for artificial intelligence (AI) in health care.

DSI leaders held face-to-face talks with decision makers at the FDA Center for Devices and Radiological Health (CDRH). They also presented and facilitated discussion on pre-market and post-market regulatory review pathways for AI at two other events with FDA officials.

Keith Dreyer, DO, PhD, the DSI’s Chief Science Officer, presented on the ACR DSI during the 2017 Medical Device Epidemiology Network’s (MDEpiNet) Annual Meeting.

MDEpinet is an FDA-launched initiative to address the needs of the medical device clinical ecosystem by strategic development of infrastructure, innovative methodological approaches for conducting initial evaluation, and creating surveillance tools that will improve medical device safety and effectiveness throughout the device life cycle.

Dreyer presented the plan, mission and objectives of the ACR DSI as an example of a platform for gathering real-world evidence to determine AI algorithm effectiveness in clinical practice. Although historically the networks in MDEpinet have been based on pathways used for traditional medical devices, the 21st Century Cures Act of 2016 clarified FDA regulations on medical software. This statutorily excluded some software functions as a medical device.

The FDA is currently developing guidance and potentially novel review pathways around Software as a Medical Device (SaMD) to establish an oversight process that promotes innovation and potentially streamlines the regulatory process.

While AI algorithms may classify as SaMD, it will remain important for the FDA to recognize the potential public safety issues that will arise around the use of AI in health care.

Additionally, the FDA is building the National Evaluation System for Health Technology (NEST) as a process to shorten the time to market for new technology health care products. Its stated purpose is to bring life-sustaining, health-promoting devices to patients more quickly by moving to more active surveillance in clinical practice. Panel discussants during the session, which included representatives from the FDA and the ONC, spoke positively about improving FDA post-market surveillance processes with ACR DSI AI algorithm monitoring tools, which employ registry reporting of real-world data.

Earlier in the week, Bibb Allen, MD, the ACR DSI’s Chief Medical Officer, met with FDA officials from the Center for Devices and Radiological Health (CDRH) in an educational session for FDA officials sponsored by General Electric. Allen presented the ACR DSI AI algorithm validation program. He discussed the opportunity for ACR DSI decentralized, multi-center validation tools to qualify as a Medical Device Development Tool (MDDT) for developers to potentially streamline their FDA pre-market review process of AI algorithms. Encouraged by these discussions about its validation process, the ACR DSI will seek further guidance on how best to propose and qualify its tools in the MDDT program.

The ACR DSI is excited about the opportunities to forge partnerships with industry and government regulators to ensure health care AI is implemented in a safe and effective manner. The ACR DSI is in the process of conducting proof of concept demonstrations to validate both its pre-market review and post-market surveillance tools and will continue to champion these tools in discussions with industry developers and government regulators.