We are delighted to share that our paper that demonstrates the need for task-based evaluation of AI-based denoising algorithms has been accepted for publication! This study shows, through an in silico imaging trial (also referred to as virtual imaging trial) that evaluation of AI-based denoising algorithms using fidelity-based metrics such as root mean square error (RMSE) and structural similarity (SSIM) can yield findings that are not consistent with evaluation of these methods on the tasks that they have been acquired for! Thus, there is an important need for task-based evaluation of AI algorithms. Further, our study shows that in silico imaging trials can provide a mechanism for conducting such evaluations.
Congratulations to all the collaborators on this study! Special congrats to Zitong, the first author on this paper.