Deep-Learning-Based Magnetic Resonance Simultaneous Multislice Imaging Using Holographic Image Decoding

We propose a new magnetic resonance simultaneous multislice imaging (SMS) by developing our previously proposed imaging method based on the principle of holography. Different amplitude modulations for each slice are applied to separate the slice images. The slice image separation is performed using a data-driven science in which U-Nets separates the slice images using the given amplitude modulation as a cue. Simulation experiments using a real-valued and an phase-varied image model revealed the relationship between the coefficients of amplitude modulation, the number of simultaneously excited slices, and image quality. In addition, we performed image separation experiments on the holographic imaging method by approximating the excited volume region as five slice images with no gaps, and showed the feasibility of image separation. This method has the potential to be a new type of SMS.