Robust water–fat separation based on deep learning model exploring multi‐echo nature of mGRE

Abstract

Purpose To design a new deep learning network for fast and accurate water–fat separation by exploring the correlations between multiple echoes in multi‐echo gradient‐recalled echo (mGRE) sequence and evaluate the generalization capabilities of the network for different echo times, field inhomogeneities, and imaging regions. Methods A new multi‐echo bidirectional convolutional residual network (MEBCRN) was designed to separate water and fat images in a fast and accurate manner for the mGRE data. This new MEBCRN network contains 2 main modules, the first 1 is the feature extraction module, which learns the correlations between consecutive echoes, and the other one is the water–fat separation module that processes the feature information extracted from the feature extraction module. The multi‐layer feature fusion (MLFF) mechanism and residual structure were adopted in the water–fat separation module to increase separation accuracy and robustness. Moreover, we trained the network using in vivo abdomen images and tested it on the abdomen, knee, and wrist images. Results The results showed that the proposed network could separate water and fat images accurately. The comparison of the proposed network and other deep learning methods shows the advantage in both quantitative metrics and robustness for different TEs, field inhomogeneities, and images acquired for various imaging regions. Conclusion The proposed network could learn the correlations between consecutive echoes and separate water and fat images effectively. The deep learning method has certain generalization capabilities for TEs and field inhomogeneity. Although the network was trained only in vivo abdomen images, it could be applied for different imaging regions.

Accelerating multi‐echo chemical shift encoded water–fat MRI using model‐guided deep learning

“…A 6-peak fat model was used with the following peak frequencies: −3.80, −3.40, −2.60, −1.94, −0.39, and 0.59 ppm. 33 The relative amplitudes of the spectral peaks are: 0.087, 0.694, 0.128, 0.004, 0.039, and 0.048, respectively.…”

“…A novel framework, which combines the power of data‐driven learning with the model‐based reconstruction, has recently been proposed to solve the general inverse problems for MRI 27–31 . Several deep learning methods using different networks have been used to improve the image quality and efficiency for water–fat MRI 32,33 . However, using deep learning to directly retrieve the water/fat images from the undersampled k‐space data in the accelerated water–fat imaging has not yet been well reported.…”

Swap-Free Fat-Water Separation in Dixon MRI using Conditional Generative Adversarial Networks

“…All the above methods were based or closely related to the U-Net architecture 33 for 2D data. A bi-directional convolutional residual network has been shown to outperform the common U-Net in multi-echo gradient recalled echo data, where the results improve when increasing the number of echo times 34 . In 35 , the authors proposed a CNN that separates fat and water using real and imaginary data from six echo times of single-slice knee and head multiecho MRI.…”

Deep learning–based quantitative susceptibility mapping (QSM) in the presence of fat using synthetically generated multi‐echo phase training data

“…The fat–water separation problem has also been addressed recently with DL methods. For example, Jafari et al compared supervised, unsupervised, and no‐training deep learning approaches to generating corrected field maps, 11 and Liu et al used multi‐echo GRE data to obtain water and fat images from in vivo data 12 …”

“…For example, Jafari et al compared supervised, unsupervised, and no-training deep learning approaches to generating corrected field maps, 11 and Liu et al used multi-echo GRE data to obtain water and fat images from in vivo data. 12 In this study, we propose a fast and automated DL method to directly reconstruct susceptibility maps from unwrapped phase maps, so that background field removal and dipole inversion are integrated, and with no need for manual background field masking. Our approach can extend the use of DL QSM reconstruction methods to regions that would otherwise suffer from chemical shift artifacts such as the neck, abdomen, pelvis, and knee.…”