Research Article
Automated Segmentation of Mass Regions in DBT Images Using a Dilated DCNN Approach
Table 1
The proposed dilated DCNN configuration and parameters.
| Layer | Type | Configuration | Dilation | Number of parameters |
| 1 | Convolutional | 3 × 3 × 1 × 32 | 1 | 2256 | Batch normalization | ReLU |
| 2 | Convolutional | 3 × 3 × 32 × 32 | 1 | 18464 | Batch normalization | ReLU |
| 3 | Convolutional | 3 × 3 × 32 × 32 | 2 | 36928 | Batch normalization | ReLU |
| 4 | Convolutional | 3 × 3 × 32 × 32 | 4 | 73856 | Batch normalization | ReLU |
| 5 | Convolutional | 3 × 3 × 32 × 32 | 8 | 147712 | Batch normalization | ReLU |
| 6 | Convolutional | 3 × 3 × 32 × 32 | 16 | 295424 | Batch normalization | ReLU |
| 7 | Convolutional | 3 × 3 × 32 × 32 | 1 | 295424 | Batch normalization | ReLU |
| 8 | Convolutional | 1 × 1 × 32 × 2 | 1 | 16384 |
|
|