Skip to main content
. 2022 Aug 11;12:886517. doi: 10.3389/fonc.2022.886517

Figure 3.

Figure 3

Hybrid-WNet. (Left) Pre-trained 2D DenseNet 121 converted into UNet, where final features, filters from 2D prediction, are 32 and concatenated into 3D DenseNet. (Right) 3D DenseNet architecture was defined as two layers, two convolution blocks in layer 0, three convolution blocks in layer 1, and eight initial filters. BN, batch normalization.