ETER-net: End to End MR Image Reconstruction Using Recurrent Neural Network Changheun Oh 1,2(B ) , Dongchan Kim 2 , Jun-Young Chung 2 , Yeji Han 2 , and HyunWook Park 1 1 Korea Advanced Institute of Science and Technology, Daejeon 34141, Republic of Korea choh@athena.kaist.ac.kr, hwpark@kaist.ac.kr 2 Gachon University, Incheon 21565, Republic of Korea Abstract. Recently, an end-to-end MR image reconstruction tech- nique, called AUTOMAP, was introduced to simplify the complicated reconstruction process of MR image and to improve the quality of reconstructed MR images using deep learning. Despite the benefits of end-to-end architecture and superior quality of reconstructed MR images, AUTOMAP suffers from the large amount of training param- eters required by multiple fully connected layers. In this work, we pro- pose a new end-to-end MR image reconstruction technique based on the recurrent neural network (RNN) architecture, which can be more effi- ciently used for magnetic resonance (MR) image reconstruction than the convolutional neural network (CNN). We modified the RNN architec- ture of ReNet for image domain data to reconstruct an MR image from k-space data by utilizing recurrent cells. The proposed network recon- structs images from the k-space data with a reduced number of param- eters compared with that of fully connected architectures. We present a quantitative evaluation of the proposed method for Cartesian trajectories using nMSE and SSIM. We also present preliminary images reconstructed from k-space data acquired in the radial trajectory. Keywords: Image reconstruction · Neural network · End to end RNN · AUTOMAP · ReNet 1 Introduction In magnetic resonance imaging (MRI), k-space data is acquired using a vari- ety of MR sequences consisting of different radiofrequency and gradient pulses. Then, by transforming the frequency information of the k-space data into spatial information, an MR image can be reconstructed. Generally, 2D or 3D Fourier transform (FT) is used for reconstruction of an MR image because the k-space is typically scanned in a Cartesian trajectory. However, relying entirely on the FT may not be sufficient for certain applications of MRI such as non-Cartesian MRI, c  Springer Nature Switzerland AG 2018 F. Knoll et al. (Eds.): MLMIR 2018, LNCS 11074, pp. 12–20, 2018. https://doi.org/10.1007/978-3-030-00129-2_2