Figure 5.

3D U-net architecture for cine MRI spatio-temporal de-aliasing. Reconstructions from undersampled cine MRI data are given as an input. The numbers on top of the blue bars denote the number of channels for each layer. The resolution for each multilevel decomposition is shown in gray on the left. Each convolutional layer is equipped with a rectified linear unit (ReLU) as non-linear activation function. The residual U-net contains a skip connection at each scale between encoder and decoder path (concat and/or addition).