Research Article
Diagnosis System of Microscopic Hyperspectral Image of Hepatobiliary Tumors Based on Convolutional Neural Network
Table 1
Convolution structure of 3D convolutional encoder.
| Number of layers | Input size | Feature number | Number of convolution kernels | Convolution kernel size | Step size | Filling method | Pooling size | Whether batch normalization | Activation function |
| 1 | | 1 | 32 | | (1,1,1) | Valid | | Yes | Relu | 2 | | 32 | 64 | | (1,1,1) | Valid | | Yes | Relu | 3 | | 64 | 128 | | (1,1,1) | Valid | | Yes | Relu | 4 | 1 | 128 | 500 | 128 | — | — | — | Yes | Relu | 5 | 1 | 500 | n | 500 | — | — | — | — | Softmax |
|
|