Leveraging the prediction uncertainty of an artificial intelligence mammography interpretation model can reduce the workload in screen reading without decreasing performance by considering the model’s interpretations only when it is confident.
Verboom et al. (Fri,) studied this question.