Deep maximum a posterior estimator for accelerated MRI reconstruction

Tingting Wang, Shengcheng Ye, Faming Fang, Guixu Zhang, Yuanyi Zheng

Abstract: Highlights•We introduce an innovative observation model/likelihood function that captures the relationship between the slice of interest and its neighboring slices and introduce a robust joint alignment and fusion module to align and fuse multi-slice input.•We incorporate an image prior that is split into HF and LF components using the Framelet transform, which allows for more dedicated regularization and enhances the constraints imposed during reconstruction.