Pitfalls in Interpretive Applications of Artificial Intelligence in Radiology

Abstract

Interpretive artificial intelligence (Al) tools are poised to change the future of radiology. However, certain pitfalls may pose particular challenges for optimal Al interpretative performance. These include anatomic variants, age-related changes, postoperative changes, medical devices, image artifacts, lack of integration of pri- or and concurrent imaging examinations and clinical information, and the satisfaction-of-search effect. Model training and development should account for such pitfalls to minimize errors and optimize interpretation accuracy. More broadly, Al algorithms should be exposed to diverse and complex training datasets to yield a holistic interpretation that considers all relevant information beyond the individual examination. Successful clinical deployment of Al tools will require that radiologist end users recognize these pitfalls and other limitations of the available models. Furthermore, developers should incorporate explainable Al techniques (e.g., heat maps) into their tools, to improve radiologists' understanding of model outputs and to enable radiologists to provide feedback for guiding continuous learning and iterative refinement. In this article, we provide an overview of common pitfalls that radiologists may encounter when using interpretive Al products in daily practice. We present how such pitfalls lead to Al errors and offer potential strategies that Al developers may use for their mitigation.