Journal of Healthcare Engineering

Research Article

Detection of COVID-19 Case from Chest CT Images Using Deformable Deep Convolutional Neural Network

Table 1

Summary of the proposed deformable deep CNN and its base normal CNN.
Deformable CNNNormal CNN
LayersOutput shapeLayersOutput shape
Input layer[150, 150, 3]Input layer[150, 150, 3]
Conv 2D[148, 148, 16]Conv2D[148, 148, 16]
Batch normalization[148, 148, 16]Batch normalization[148, 148, 16]
Max pooling 2D[74, 74, 16]Max pooling 2D[74, 74, 16]
Deform conv 2D[74, 74, 16]Conv 2D[74, 74, 16]
Conv 2D[72, 72, 32]Conv2D[72, 72, 32]
Max pooling 2D[36, 36, 32]Max pooling 2D[36, 36, 32]
Deform conv 2D[36, 36, 32]Conv 2D[36, 36, 32]
Conv 2D[34, 34, 64]Conv2D[34, 34, 64]
Max pooling 2D[17, 17, 64]Max pooling 2D[17, 17, 64]
Dropout[17, 17, 64]Dropout[17, 17, 64]
Flatten[18496]Flatten[18496]
Dense[256]Dense[256]
Dropout[256]Dropout[256]
Dense[2]Dense[2]
Total parameters 4,764,098Total parameters 4,761,154
The bold terms are the main focus of our proposed deformable CNN architecture.