Table 7 Structures of the deep networks used for comparative analysis. (MaxPool(3) = 3 × 3 pooling layer)
From: dSPIC: a deep SPECT image classification network for automated multi-disease, multi-lesion diagnosis
| SI_CLF | SI_CLF + AM | SI_CLF + RM | SI_CLF + AM + RM | AlexNet |
|---|---|---|---|---|
| 11 × 11, 64, S = 4, P = 2 | 11 × 11, 64, S = 4, P = 2 | 11 × 11, 64, S = 4, P = 2 | 11 × 11, 64, S = 4, P = 2 | 11 × 11, 64, S = 4, P = 2 |
| MaxPool(3), S = 2 | MaxPool(3), S = 2 | MaxPool(3), S = 2 | MaxPool(3), S = 2 | MaxPool(3), S = 2 |
| 7 × 7, 128, S = 1, P = 2 | AM | RM | RM | 5 × 5, 256, S = 1, P = 2 |
| MaxPool(3), S = 2 | 7 × 7, 128, S = 1, P = 2 | 7 × 7, 128, S = 1, P = 2 | AM | MaxPool(3), S = 2 |
| 5 × 5, 128, S = 1, P = 3 | MaxPool(3), S = 2 | MaxPool(3), S = 2 | 7 × 7, 128, S = 1, P = 2 | 3 × 3, 384, S = 1, P = 3 |
| 3 × 3, 256, S = 1, P = 3 | 5 × 5, 128, S = 1, P = 3 | 5 × 5, 128, S = 1, P = 3 | MaxPool(3), S = 2 | 3 × 3, 384, S = 1, P = 3 |
| 3 × 3, 256, S = 1, P = 1 | 3 × 3, 256, S = 1, P = 3 | 3 × 3, 256, S = 1, P = 3 | 5 × 5, 128, S = 1, P = 3 | FC (1024) |
| MaxPool(3), S = 2 | 3 × 3, 256, S = 1, P = 1 | 3 × 3, 256, S = 1, P = 1 | 3 × 3, 256, S = 1, P = 3 | FC (1024) |
| FC (1024) | AM | MaxPool(3), S = 2 | 3 × 3, 256, S = 1, P = 1 | Softmax (3) |
| FC (1024) | MaxPool(3), S = 2 | FC (1024) | AM | |
| Softmax (3) | FC (1024) | FC (1024) | MaxPool(3), S = 2 | |
| FC (1024) | Softmax (3) | FC (1024) | ||
| Softmax (3) | FC (1024) | |||
| Softmax (3) |