Super-resolution Y-Net for simultaneous 1H MRF/23Na MRI

Motivation: To improve resolution for translating sodium MRI into clinical practice. Goal(s): Develop a super-resolution neural network for brain sodium images. Approach: A cascaded Y-Net is proposed to generate high-resolution sodium images from simultaneously acquired 1H MRF/23Na MRI data. Human brain images from 8 healthy subjects were used for training and validation (154), and testing (22). Results: The generated high-resolution sodium images from the Y-Net showed a structural similarity index measure (SSIM) of 0.935, a RMSE=0.034 and a PSNR=28.8 compared with the ground truth. Impact: We introduce a Y-Net super-resolution neural network that generates high-resolution sodium images from simultaneously acquired 1H MRF/23Na MRI data.