Retrospective respiratory motion correction in cardiac cine MRI reconstruction using adversarial autoencoder and unsupervised learning

Abstract

The aim of this study was to develop a deep neural network for respiratory motion compensation in free‐breathing cine MRI and evaluate its performance. An adversarial autoencoder network was trained using unpaired training data from healthy volunteers and patients who underwent clinically indicated cardiac MRI examinations. A U‐net structure was used for the encoder and decoder parts of the network and the code space was regularized by an adversarial objective. The autoencoder learns the identity map for the free‐breathing motion‐corrupted images and preserves the structural content of the images, while the discriminator, which interacts with the output of the encoder, forces the encoder to remove motion artifacts. The network was first evaluated based on data that were artificially corrupted with simulated rigid motion with regard to motion‐correction accuracy and the presence of any artificially created structures. Subsequently, to demonstrate the feasibility of the proposed approach in vivo, our network was trained on respiratory motion‐corrupted images in an unpaired manner and was tested on volunteer and patient data. In the simulation study, mean structural similarity index scores for the synthesized motion‐corrupted images and motion‐corrected images were 0.76 and 0.93 (out of 1), respectively. The proposed method increased the Tenengrad focus measure of the motion‐corrupted images by 12% in the simulation study and by 7% in the in vivo study. The average overall subjective image quality scores for the motion‐corrupted images, motion‐corrected images and breath‐held images were 2.5, 3.5 and 4.1 (out of 5.0), respectively. Nonparametric‐paired comparisons showed that there was significant difference between the image quality scores of the motion‐corrupted and breath‐held images (P < .05); however, after correction there was no significant difference between the image quality scores of the motion‐corrected and breath‐held images. This feasibility study demonstrates the potential of an adversarial autoencoder network for correcting respiratory motion‐related image artifacts without requiring paired data.

Temporally aware volumetric generative adversarial network‐based MR image reconstruction with simultaneous respiratory motion compensation: Initial feasibility in 3D dynamic cine cardiac MRI

“…We found TAV‐GAN as a promising technique for reconstructing highly undersampled and respiratory motion‐corrupted 4D data sets. Several previous deep learning–based image‐reconstruction techniques, in particular, the GAN‐based approach, are focused on undersampled data recovery 21,43,48,66 or motion compensation 49‐53,67 . In this work, we proposed the TAV‐GAN for simultaneous undersampled k‐space data recovery and respiratory‐motion compensation.…”

“…Convolutional neural networks (CNNs) and generative adversarial networks (GANs) have shown promises for MR image reconstruction 21‐48 and motion correction 49‐53 . Three‐dimensional (3D) CNNs or 2D convolutional recurrent neural networks (CRNNs) have been proposed to exploit the spatiotemporal information in 2D dynamic MRI 24,33,35,44 .…”

“…20,21 Convolutional neural networks (CNNs) and generative adversarial networks (GANs) have shown promises for MR image reconstruction and motion correction. [49][50][51][52][53] Threedimensional (3D) CNNs or 2D convolutional recurrent neural networks (CRNNs) have been proposed to exploit the spatiotemporal information in 2D dynamic MRI. 24,33,35, 44 Qin et al proposed a novel 2D CRNN framework to reconstruct high-quality 2D cardiac MR images from undersampled kspace data (6X-11X) by exploiting the temporal redundancy and unrolling the traditional optimization algorithms.…”

Cardiac MR: From Theory to Practice

“…The goal of the generator network is to fool the discriminator network to generate images that look like real motion-corrected images. Another method of retrospective motion correction in CMR with adversarial training is proposed in Ghodrati et al ( 354 ). Here, a Variational Autoencoders is trained on healthy subjects and patients with suspected cardiovascular disease to remove respiratory motion.…”

Dual-domain unsupervised network for removing motion artifact related to Gadoxetic acid-enhanced MRI

“…Armanious et al proposed Cycle-MedGAN [9] and Cycle-MedGAN V2.0 [10], and they mainly focused on reducing rigid artifacts in brain MRI and achieved promising results. Ghodrati et al [11] presented a framework based on the adversarial autoencoder for respiratory motion correction. Their method can effectively reduce blurring and artifacts associated with respiratory motion.…”