Deep-learning reconstruction for low-field MRI

This research explores deep-learning-based reconstruction for low-field MRI, addressing diagnostic-quality challenges and accessibility in underserved regions. Utilizing 0.1T MR scans, a Residual U-Net improves image quality through joint magnitude-phase training and optimized metrics (SSIM, PSNR). Prospective and retrospective evaluations demonstrate efficient reconstruction performance despite limited datasets. Findings highlight advancements in low-field MRI imaging, providing insights...