Figure 1:

(a) Schematic of recurrent Registration Network (RRN), where convolutional (Conv) layers in the encoder are combined with 3D-CLSTM. (b) Recurrent Segmentation Network (RSN) uses a Unet-3D backbone with 3D-CLSTM placed after convolutional blocks in the encoder layers. (c) ProRSeg combines RRN and RSN. The unrolled representation showing CLSTM in the encoder layers for progressively refining the registration and segmentation are shown. RSN combines $x_t$ with the progressively aligned images $x_m^{i=1,\dots N}$ and segmentations $y_m^{i=1,\dots N}$ produced by RRN as inputs to its CLSTMs to generate segmentation $y_t$ in N steps.