As the integration of artificial intelligence-enabled devices and software into clinical practice gains momentum, a new paper delves into whether AI lives up to its reputation as a trusted aid for radiologists.
Published in American Journal of RoentgenologyThe authors of a follow-up to an earlier study published in Nature Medicine asserted that AI can both help and hinder radiologists.
“There is growing enthusiasm for AI assistance in radiologists' daily workflow. Studies to date suggest that AI-assisted detection could be of great value in CXR interpretation,” said co-authors Angela Udonwo, MD, PhD, of the Lewis Katz School of Medicine at Temple University, and Farouk Dako, MD, PhD, of the Department of Radiology at the Perelman University Hospital of the University of Pennsylvania. “This study highlights the limitations of AI in improving radiologists' performance in CXR reading, particularly the variability and unpredictability of its impact.”
In the first study, the researchers analyzed the performance of 140 radiologists who read chest x-rays with and without the help of AI. Factors such as experience and skill level, AI accuracy, subspecialty, and history of using AI tools were all taken into account when evaluating the radiologists' performance on 15 different diagnostic tasks related to reading chest x-rays.
Experts have concluded that in many cases, radiologists' performance is only marginally improved with AI assistance. Additionally, the benefits of using AI are not consistent across all readers, with radiologists with the least experience using AI reporting greater benefit from AI assistance than radiologists with more experience using AI. It seemed like they didn't receive much attention.
