Statistical Considerations and Challenges for Pivotal Clinical Studies of Artificial Intelligence Medical Tests for Widespread Use: Opportunities for Inter-Disciplinary Collaboration

Abstract

The application of Artificial Intelligence to medical testing has received much attention in recent years, as evidenced by the flurry of published studies describing Artificial Intelligence software developed to solve problems in medical testing. While this recent activity is exciting, developed Artificial Intelligence medical tests ultimately can only be considered as candidates for widespread use if these tests demonstrate good performance in pivotal clinical studies. What are pivotal clinical studies for Artificial Intelligence medical tests aimed for widespread use? What are some of the major considerations and challenges for assessing performance of these tests in this context? What are some of the outstanding areas where statisticians, in collaboration with professionals outside the statistical community, could help in this endeavor? This article addresses these questions. This article is meant to appeal to a broad audience with varying levels of statistical and medical testing knowledge so that inter-disciplinary collaboration could be enhanced.