Skip to main content
. 2020 Feb 25;7:17. doi: 10.3389/fcvm.2020.00017

Figure 7.

Figure 7

Different 2D and 3D deep learning-based approaches for radial undersampling artifacts reduction (post-processing) presented in Kofler et al. (75). (A) 2D U-net for frame-to-frame mapping. (B) 2D U-net for sequence-to-sequence mapping with cardiac phases aligned along the channel dimension. (C) 3D U-net for sequence-to-sequence mapping with 3D convolutional kernels. (D) 2D U-net for recovery of two-dimensional spatio-temporal images.