Dose_prediction_han_vmat

<h1 style="border-bottom: 2px solid black; font-size: 100px;" align="center"> HDUNet VMAT </h1>

Trained by Margerie Huet Dastarac . <br> Training date: 06/06/2023 .

1. Task Description

Prediction of the dose distribution with VMAT scanning treatment.

2. Model

2.1. Architecture

<img width="100%" src="https://cdn-uploads.huggingface.co/production/uploads/65c9dbefd6cbf9dfed67367e/xGGUE5spRY6tar5R4VeKX.png" alt="error while loading image">

Figure 1: HDUNet VMAT architecture

2.2. Input

<ul> <li> CT: 3D float matrix</li> <li> Target volumes contours and prescription: 3D float matrices: pixel value set with prescription in different targt volumes</li> <li> Organs at risks contours: 3D boolean matrices, one per considered organ</li> </ul>

2.3. Output

<ul> <li> DOSE: 3D float matrix</li> </ul>

2.4 Training details

<ul> <li> Number of epoch: 400 </li> <li> Loss function: MSE loss </li> <li> Optimizer: AdamW </li> <li> Learning Rate: 0.0001 </li> <li> Dropout: No </li> <li> Patch size in voxels: (128,128,128) </li> <li> Data augmentation used: <ul> <li> RandCrop</li> </ul> </li> </ul>

3. Dataset

<ul> <li> Location: Oropharynx </li> <li> Training set size: 57 </li> <li> Resolution in mm: 3x3x3 </li> </ul>

Performance

<img width="50%" src="https://huggingface.co/AI4MIRO/Dose_prediction_han_vmat/resolve/main/images/Figure4_vmat.JPG">