HDUNet VMAT
Trained by Margerie Huet Dastarac .
Training date: 06/06/2023 .
1. Task Description
Prediction of the dose distribution with VMAT scanning treatment.
2. Model
2.1. Architecture
Figure 1: HDUNet VMAT architecture
2.2. Input
- CT: 3D float matrix
- Target volumes contours and prescription: 3D float matrices: pixel value set with prescription in different targt volumes
- Organs at risks contours: 3D boolean matrices, one per considered organ
2.3. Output
- DOSE: 3D float matrix
2.4 Training details
- Number of epoch: 400
- Loss function: MSE loss
- Optimizer: AdamW
- Learning Rate: 0.0001
- Dropout: No
- Patch size in voxels: (128,128,128)
- Data augmentation used:
- RandCrop
3. Dataset
- Location: Oropharynx
- Training set size: 57
- Resolution in mm: 3x3x3
Performance
