Validation of a Breakthrough AI-Guided Echocardiography System: Overcoming Challenges in Testing AI-Based Medical Device

• Activity Number: 378 - Study Design and Statistical Challenges for AI/ML Based Medical Tests
• Type: Topic-Contributed
• Date/Time: Thursday, August 12, 2021 : 12:00 PM to 1:50 PM
• Sponsor: Section on Medical Devices and Diagnostics
• Author(s): Ha Hong* and Samuel Surette and Yngvil Thomas and Charles Cadieu and Ali Chaudhry and Randolph Martin
• Companies: Caption Health, Inc. and Caption Health, Inc. and Caption Health, Inc. and Caption Health, Inc. and Caption Health, Inc. and Caption Health, Inc.
• Keywords: echocardiography; medical device; clinical trial design; artificial intelligence; deep learning

Abstract

Cardiac ultrasound is a primary diagnostic tool for detection and diagnosis of cardiovascular disease, the number one killer in the United States as well as worldwide. While providing rich clinical information, the ultrasound exam is difficult to perform and requires skilled experts. We have applied deep learning to make an AI-guided echocardiography system, a transformational new ultrasound technology that enables healthcare providers—even those without prior experience—to perform ultrasound exams quickly and accurately by providing expert guidance, automated quality assessment, and intelligent interpretation. In this presentation, we focus on how we developed the technology and validated its safety and effectiveness. We introduce a system/procedure to improve AI algorithms in a more safe and structured way. We describe how we overcame statistical challenges in testing this AI-based medical device that interacts with the operator in real time based on the images from the patient. We then discuss the outcome of the clinical trial that is designed by taking this AI-operator-patient interaction into account.