Multi-Channel Grouped CNN-Based Image Reconstruction For Reduced Sampled Complex MR Signal.

Applying Compressed Sensing to Magnetic Resonance Image Reconstruction using CNN (CNN-CS) has attracted much attention. CNN-CS enables us to reconstruct an image quickly with better quality than traditional mathematical iterative Compressed Sensing methods (CS-MRI). While many CNN-CS methods have been proposed, they have problems recovering high frequency components and removing aliasing artifacts. In this study, we proposed a novel Transformed Image Domain CNN-CS using multi-channel grouped CNN-based image reconstruction using the Fresnel transform (eFRBAS transform). Experimental results showed that the proposed method was able to predict an artifact-free image better than other methods, especially for a 20-30% low sampling rate.