Device and Method for Reconstructing Undersampled Magnetic Resonance Image

摘要

The present invention relates to an under-sampled magnetic resonance imaging (MRI) image reconfiguration apparatus and a method thereof. The under-sampled MRI image reconfiguration apparatus includes: a K-space network for reconfiguring K-space images of the under-sampled MRI images; a first correction unit for correcting K-space reconfiguration images being output from the K-space network; an image network which executes a reconfiguration operation for removing artifacts by receiving the MRI images corresponding to the images corrected by the first correction unit as an input; and a second correction unit which corrects the reconfigured MRI images being output from the image network. The K-space network uses the K-space images of normally sampled MRI images and K-space images of the under-sampled MRI images to learn to convert the K-space images of the under-sampled MRI images in a form similar to that of the normally sampled MRI images. The image network uses the normally sampled MRI images and the MRI images related to the output images from the K-space network to learn to convert the MRI images related to the output images of the K-space network in a form similar to that of the normally sampled MRI images. The apparatus and the method can use a deep learning network to more efficiently remove the artifacts occurring during the reconfiguration of the under-sampled MRI images.