1388. FDA Approval for Artificial Intelligence Applications in Healthcare

Authors: Rui Song; Eric Oermann, MD; Remi Kessler, BA; Kurt Yaeger, MD (Brooklyn, NY)

Introduction: Over the past several years, Artificial Intelligence (AI) has become a powerful adjunct to improve clinician accuracy. As such, the US Food and Drug Administration has focused intently on regulating AI solutions for healthcare. The introduction of the FDA de novo pathway for medical device approval has facilitated market entry of clinical software applications. In this abstract, we present current FDA approved AI healthcare applications and discuss the future of FDA regulations on AI-based technologies. Methods: A review of artificial intelligence enabled software and devices was conducted by searching keyword terms artificial intelligence, machine learning, algorithm in PubMed and the official FDA website. Results: Our search resulted in only 4 approved AI-based healthcare aids; all but one was approved via the FDA’s de novo pathway. Viz AI’s ContaCT software, approved in February 2018, interprets CT angiography for large vessel occlusion. IDx-Dr by IDx LLC, approved in April 2018, aids diagnosis of diabetic retinopathy on fundoscopy. Imagen’s OsteoDefect, approved in May 2018, interprets 2D X-ray images for distal radial fractures. Aidoc’s BriefCase software, approved by the FDA’s 510k program, is based on Viz AI’s ContaCT and identifies intracerebral hemorrhage on head CT. Conclusion: Healthcare software applications that are used as medical devices require FDA approval. The FDA De Novo pathway was developed for medical devices for which there may be no substantial equivalent and has been used frequently for AI-based clinical software. Four AI-based software applications are approved currently by the FDA, all related to image processing and interpretation. The current FDA de novo program is rigorous, requiring proof of concept for approval. However, future applications may be cleared through the 510k pathway based on existing AI technology. As AI is refined and regulatory efforts become clearer, we expect to see more frequent FDA approval for AI-based healthcare solutions.