Deep learning-based dose prediction for low-energy electron beam superficial radiotherapy

Area of Science:

• Medical Physics
• Radiotherapy
• Computational Biology

Background:

• Accurate surface dose calculation is critical in superficial low-energy electron beam radiotherapy.
• Traditional Monte Carlo (MC) simulations offer precision but are computationally intensive and slow.

Purpose of the Study:

• To enhance the speed and accuracy of surface dose calculations in radiotherapy.
• To develop a deep learning model for rapid dose distribution prediction.

Main Methods:

• Combined MC simulations with a cascaded 3D U-Net (C3D) deep learning model.
• Simulated low-energy electron beams using DOSXYZnrc for six body sites.
• Trained the C3D model on generated dose distributions for rapid prediction.

Main Results:

• The C3D model achieved high accuracy with a minimum Gamma pass rate of 92.09%.
• Dose predictions were completed in 0.42 seconds, approximately 140,000 times faster than MC simulations.
• C3D outperformed other deep learning models in accuracy and robustness across six anatomical regions.

Conclusions:

• Deep learning integration with MC simulations significantly improves efficiency in surface dose calculations.
• The C3D model enables rapid and accurate dose predictions for efficient radiotherapy treatment planning.