Learning a variational network for reconstruction of accelerated MRI data

TLDR

The study aims to enable rapid, high‑quality reconstruction of accelerated multi‑coil MRI by learning a variational network that merges variational model structure with deep learning. The authors embed a generalized compressed‑sensing variational model in an unrolled gradient‑descent scheme, learn all parameters offline, and apply the resulting network online to accelerate knee MRI with various sampling patterns. The variational network outperforms conventional reconstruction methods, preserves natural image appearance and unseen pathologies, and achieves 193 ms reconstruction time on a single GPU, enabling seamless clinical integration. Published in Magn Reson Med 79:3055‑3071 (2018); © 2017 International Society for Magnetic Resonance in Medicine.

Abstract

To allow fast and high-quality reconstruction of clinical accelerated multi-coil MR data by learning a variational network that combines the mathematical structure of variational models with deep learning.Generalized compressed sensing reconstruction formulated as a variational model is embedded in an unrolled gradient descent scheme. All parameters of this formulation, including the prior model defined by filter kernels and activation functions as well as the data term weights, are learned during an offline training procedure. The learned model can then be applied online to previously unseen data.The variational network approach is evaluated on a clinical knee imaging protocol for different acceleration factors and sampling patterns using retrospectively and prospectively undersampled data. The variational network reconstructions outperform standard reconstruction algorithms, verified by quantitative error measures and a clinical reader study for regular sampling and acceleration factor 4.Variational network reconstructions preserve the natural appearance of MR images as well as pathologies that were not included in the training data set. Due to its high computational performance, that is, reconstruction time of 193 ms on a single graphics card, and the omission of parameter tuning once the network is trained, this new approach to image reconstruction can easily be integrated into clinical workflow. Magn Reson Med 79:3055-3071, 2018. © 2017 International Society for Magnetic Resonance in Medicine.