Skip to main content

Table 2 Details of architecture

From: A convolutional neural network-based system to classify patients using FDG PET/CT examinations

Layer Filter Size Stride Repeat count Output Size
Input     (224, 224, 3)
Convolutional (7, 7) (2, 2) 1 (112, 112, 64)
Max pooling (3, 3) (2, 2) 1 (56, 56, 64)
Residual 1 (3 × 3, 64)
(3 × 3, 64)
(1, 1) 3 (56, 56, 64)
Residual 2 (3 × 3, 128)
(3 × 3, 128)
(2, 2) 4 (28, 28, 128)
Residual 3 (3 × 3, 256)
(3 × 3, 256)
(2, 2) 6 (14, 14, 256)
Residual 4 (3 × 3, 512)
(3 × 3, 512)
(2, 2) 3 (7, 7, 512)
Average pooling (7, 7) (1, 1) 1 (1, 1, 1024)
Fully connected     (3)
  1. “Residual” contains the following structure. “1. Convolutional layer1, 2. Batch normalization1, 3. Activation layer1 (ReLU), 4. Convolutional layer2, 5. Batch normalization2, 6. Merge layer (Add), 7. Activation layer2 (ReLU)”