On Reference-Based Image Quality Assessment in Medical Image Reconstruction: Potential Pitfalls and Possible Solutions
Chin‐Cheng Chan, Jiayang Wang, Tashfa Nadeem, Justin P. Haldar · 2023
The development of new computational image reconstruction methods is heavily dependent on the ability to perform accurate image quality assessment. In recent years, the prevailing paradigm has evolved towards assessing reconstruction performance using quantitative error metrics (such as mean-squared error and structural similarity) that are calculated with respect to a database of gold-standard images. While this approach does provide useful insights, it also has some major limitations. In this paper, we review some of our group's recent work on this topic. We bring attention to the fact that popular image quality assessment methods can be disconnected from important characteristics like image resolution, and we also describe how the prevailing performance assessment approach can lead to the incorrect ranking of different image reconstruction methods. We also review potential solutions that our group has developed to address these problems.