Lancet publishes Dewey's thoughts on AI

Radiologists need to address three key issues in order for their field to flourish, and artificial intelligence (AI) can help them achieve that goal, according to a letter from Dr. Marc Dewey from Charité-Universitätsmedizin Berlin published in the August issue of the Lancet.

In the letter, Dewey discussed the following three concerns in the practice of radiology:

• Overuse of services due to subjective decisions about tests
• Inconsistent acquisition and analysis of images
• Vague descriptions in reports.

Improvements in decision-making, along with the use of validated clinical predictions for disease risk, would cut down on the overuse of certain services, he wrote. AI could also help overcome some of these challenges. For example, the integration of AI with human analysis has the potential to increase consistency in image analysis and minimize errors.

Hypothetically, this could one day lead to the development of bionic radiologists that combine the consistency of automation with human perception, Dewey added. They could generate structured radiology reports that might reduce missed-care opportunities and bridge the gap between evidence-based and personalized treatment recommendations.

Dewey's full analysis was presented in a plenary lecture at ECR 2018.