Table 2 Selected hyper parameters for training and inference
Parameter | Training | Inference |
|---|---|---|
Window size | 320, 320, 320 | 400, 400, 400 |
Activation function | PRelu | |
Loss type | DicePlusXEnt | |
Normalization | Histogram | Histogram |
Window sampling | Weighted | Weighted |
Volume padding | 24, 24, 24 | 0, 0, 0 |
Learning rate | 0.0001 | |
Random flipping axes | (0, 2) | |
Elastic transformation | True | |
Rotation angle | (-10, 10) | |
Scaling percentage | (-10, 10) |