Table 2 More details of deep learning method in COPD identification and stage, including neural network architectures, number of layers, and activation functions
From: Artificial intelligence in COPD CT images: identification, staging, and quantitation
Team | Reference | Neural network architecture | Number of layers, | Activation functions |
|---|---|---|---|---|
González et al. | [66] | CNN | Three conv and max pooling | Rectified Linear (ReLU) |
Xu et al. | [70] | AlexNet | Five conv and three max pooling | ReLU |
Du et al. | [71] | Self-designed | Some conv and max pooling | Leaky ReLU |
Tang et al. | [72] | ResNet 152 | 152 | - |
Ho et al. | [73] | 3D CNN-Naive | Three conv and max pooling | ReLU |
Zhang et al. | [78] | DenseNet-201 | 201 | ReLU |
Wu et al. | [37] | ResNet-26 | 26 | ReLU |
Sun et al. | [38] | ResNet-18 | 18 | ReLU |
Wu et al. | [32] | VGG-16 | 16 | ReLU |
Almeida et al. | [82] | 3D ResNet-34 | 34 | - |
Xue et al. | [84] | Resnet-50 | 50 | ReLU |
Yu et al. | [87] | Loc-CondConv | - | Sigmoid |