CT-MRI Image Reconstruction with Mask-enhanced Dual-Dictionary Learning

In this work, a mask-enhanced dual-dictionary learning method is proposed to solve the problem of hybrid image reconstruction in magnetic resonance imaging (MRI) and X-ray computed tomography (CT). One dictionary trained from MRI data as a high quality dictionary used to generate a base MRI, and the other dictionary trained from CT data as a low quality dictionary to get the sparse representation. Experiments with real images were performed to evaluate the proposed method. The results reconstructed using the proposed method shows improved MRI image quality compared with the reconstructions using dictionary learning (DLMRI).