Figure 3
From: Deep radiomics-based survival prediction in patients with chronic obstructive pulmonary disease
Convolutional neural network (CNN) model architecture. CNN architecture for classification of six-minute walk distance test results (> 440 m or not) to extract deep features. Each of the 11 selected slices was trained separately using the same architecture, which was designed to have five convolution blocks and one fully connected layer. Deep features were obtained by normalizing the information of the last fully connected layer (yellow). Each convolution block consisted of a convolutional layer followed by batch normalization (BatchNorm), rectified linear unit activation (Relu), and max-pooling.
