期刊
IEEE TRANSACTIONS ON MEDICAL IMAGING
卷 41, 期 7, 页码 1625-1638出版社
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
DOI: 10.1109/TMI.2022.3144619
关键词
Image reconstruction; Magnetic resonance imaging; Three-dimensional displays; Neural networks; Optimization; Sensitivity; Iterative methods; MRI; image reconstruction; deep learning; non-cartesian; unrolled networks
类别
资金
- High Performance Computing (HPC) Resources of Institut du Developpement et des Ressources en Informatique Scientifique [Institute for the Development and Resources in Scientific Computing (IDRIS)] through Grand Equipement National de Calcul Intensif [Large [2021-AD011011153]
- Cross-Disciplinary Program on Numerical Simulation of Comissariat a l'Energie Atomique et aux energies renouvelables [Atomic Energy and Renewable Energies Commission (CEA)]
This study explores the potential of deep learning in non-Cartesian acquisition settings and validates the performance of the NC-PDNet, a density-compensated unrolled neural network. The experiments demonstrate that the NC-PDNet outperforms baseline models both visually and quantitatively in various settings.
Deep Learning has become a very promising avenue for magnetic resonance image (MRI) reconstruction. In this work, we explore the potential of unrolled networks for non-Cartesian acquisition settings. We design the NC-PDNet (Non-Cartesian Primal Dual Netwok), the first density-compensated (DCp) unrolled neural network, and validate the need for its key components via an ablation study. Moreover, we conduct some generalizability experiments to test this network in out-of-distribution settings, for example training on knee data and validating on brain data. The results show that NC-PDNet outperforms baseline (U-Net, Deep image prior) models both visually and quantitatively in all settings. In particular, in the 2D multi-coil acquisition scenario, the NC-PDNet provides up to a 1.2 dB improvement in peak signal-to-noise ratio (PSNR) over baseline networks, while also allowing a gain of at least 1dB in PSNR in generalization settings. We provide the open-source implementation of NC-PDNet, and in particular the Non-uniform Fourier Transform in TensorFlow, tested on 2D multi-coil and 3D single-coil k-space data.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据