Figure 2.

Deep neural network architectures used to generate individually adaptive template for 3D amyloid PET images. (a) CAE, which was also used as the generator in the GAN. (b) Convolutional neural network used as the discriminator in the GAN. (c) Structure of GAN. Each red orange box represents a 3D strided convolutional kernel, where s is the size of the stride and k is the size of the kernels. Each blue box represents batch normalization combined with activation function such as leaky‐ReLU or ELU. The green box shows the fully connected layer with the number of units. The purple box represents the 3D transposed convolutional layer (deconvolution) with two stride and kernel size three. The generator has the architecture of the residual box, as shown in the figure [Color figure can be viewed at http://wileyonlinelibrary.com]