Assessing MR image reconstruction quality using the Fourier Radial Error Spectrum plot

This work introduces the Fourier radial error spectrum plot (ESP) as a novel approach to quantifying the quality of reconstructed MR images. While conventional error metrics such as normalized root mean squared error (NRMSE) or structural similarity (SSIM) are widely used, they are simple scalar-measures that only provide one-dimensional insight into image quality. In contrast, ESP describes reconstruction quality with a spectrum that provides a quantitative evaluation of image quality at every spatial resolution scale. Our results show that ESP provides more comprehensive information than conventional error metrics, and can guide the design of new and improved image reconstruction approaches.