Zero-Shot Simultaneous Multislice Reconstruction Based on Pretrained Generative Image Priors

Kexin Yang, Shoujin Huang, Guanxiong Luo, Lifeng Mei, Jingyu Li

Keywords: Simultaneous Multislice Imaging, Deep Learning, Generative Image Priors

Abstract:

In this study, we combine the readout-concatenation framework with generative image priors to achieve simultaneous multislice imaging (SMS) reconstruction. The results show that generative image priors have better generalization than supervised deep learning methods
such as VarNet, and it can be processed flexibly handle arbitrary slice aliasing patterns with in-plane acceleration.

Submission Number: 112