Figure 2
A diagram of a 3D convolutional neural network showing the process of feature extraction from a 3D MRI image. The input image of size 256 by 256 by 128 is processed through six convolutional layers. Each layer applies 3D convolutions with ReLU activation and pooling operations, reducing the dimensionality and extracting features. The first layer produces 32 feature maps of size 256 by 256 by 128, followed by pooling. The second layer produces 32 feature maps of size 128 by 128 by 64. The third layer produces 64 feature maps of size 128 by 128 by 64. The fourth layer produces 64 feature maps of size 64 by 64 by 32. The final layer produces 1024 feature maps of size 8 by 8 by 4, resulting in 262144 features.3D convolutional neural network