DoseNet: a volumetric dose prediction algorithm using 3D fully-convolutional neural networks.

Area of Science:

• Medical Physics
• Radiotherapy
• Artificial Intelligence in Medicine

Background:

• Accurate radiation dose prediction is crucial for effective prostate stereotactic body radiotherapy (SBRT).
• Existing methods like U-Net and fully connected networks have limitations in non-coplanar dose prediction.
• Novel deep learning approaches are needed to improve computational efficiency and prediction accuracy.

Purpose of the Study:

• To demonstrate the feasibility and performance of a novel fully-convolutional volumetric dose prediction neural network (DoseNet).
• To evaluate DoseNet as a superior alternative for non-coplanar SBRT prostate dose prediction compared to existing models.
• To assess the computational efficiency and accuracy of DoseNet in predicting radiation dose distributions.

Main Methods:

• DoseNet, a novel fully-convolutional neural network utilizing 3D convolutions and deconvolution, was developed.
• The network was implemented on GPUs and trained using a 3-phase learning protocol on 151 prostate SBRT patients.
• Dosimetric quality was assessed by comparing predicted doses with delivered doses using conformity and heterogeneity indices.

Main Results:

• DoseNet demonstrated superior performance compared to U-Net and fully connected methods for prostate SBRT dose prediction.
• The model achieved accurate volumetric dose predictions, outperforming existing methods in key dosimetric parameters.
• DoseNet exhibited high computational efficiency, requiring approximately 0.83 seconds for prediction on a 128x128x64 voxel image.

Conclusions:

• DoseNet is a feasible and effective tool for accurate volumetric dose prediction in non-coplanar SBRT prostate treatments.
• The proposed neural network offers a significant improvement over current methods, enhancing treatment planning capabilities.
• DoseNet provides a computationally efficient solution for predicting radiation dose distributions in complex radiotherapy cases.