“…However, imaging algorithms, including those based on DL, are often evaluated using figures of merit (FoMs) that are not explicitly designed to measure clinical task performance ( 11 ). Recent studies conducted specifically in the context of evaluating image-denoising algorithms showed that task-agnostic FoMs may yield interpretations that are inconsistent with evaluation on clinical tasks ( 13 – 17 ). For example, in Yu et al.…”