Super-Resolution for MRI Using Phase-Scrambling Fourier Transform Imaging and Unrolling Model-Based Network

Phase-scrambling Fourier transform imaging is an imaging technique for MRI and has useful features not found in Fourier transform imaging techniques generally used in MRI. In our previous study, an MR super-resolution method using a phase-scrambling Fourier transform signal was proposed. The previous method had two problems: a small improvement in resolution in the center regions of the reconstructed image and performance limitations in the iterative reconstruction. To solve these problems, here we propose an MR super-resolution method using a phase-scrambling Fourier transform signal and an unrolling model-based network. The proposed method utilizes Generic-ADMM-Net as the unrolling model-based network because MR super-resolution can be regarded as compressed sensing with image reconstruction from band-limited signals. To verify the effectiveness of the proposed method, we conducted simulation experiments. In the experiments, in addition to quantitative evaluation of the resolution improvement using numerical phantom images and reconstruction experiments using real-valued images, to evaluate the practical use of our method, we also conducted reconstruction experiments using noise-added images and complex images with phase components. As a result, it was confirmed that the images reconstructed from a Phase-Scrambling Fourier Transform (PSFT) signal had higher quality and better resolution improvement than those reconstructed from an FT signal. Moreover, the reconstruction performance of the proposed method outperformed that of the previous method, and the resolution of the center regions of the reconstructed image was improved. Furthermore, the results of practical experiments showed that the proposed method had a noise reduction effect and can be applied to complex images with phase components.