Skip to main content
. Author manuscript; available in PMC: 2023 Sep 1.
Published in final edited form as: J Neuroimaging. 2022 Apr 17;32(5):968–976. doi: 10.1111/jon.12997

Figure 2.

Figure 2.

IV-Net our Deep Neural Network automated method is a 5-layer network (Top) with 3 different architectural blocks (Bottom). In the encoder layers (Blue) densely connected dilated convolutions (Dilated Convs) are combined with Group Normalization (Group Norm), a Bottleneck layer and a scaling convolution (Scaling Conv). In the central layers (Grey) convolutions (3×3×3 Convs) are combined with Group Norm and a dropout layer (Dropout) before a Scaling Conv. The decoder layers (Orange) include Attention Gates. Skip connections (dashed arrows) span the network.