Multi-model study of fast VMAT segment dose calculation with deep learning

Key Points

• This research aims to evaluate the accuracy and efficiency of deep learning methods for photon dose calculation in radiotherapy.
• Utilized multi-model deep learning techniques, including BEV- and patient-coordinate methods.
• Compared performance of CNN-ConvLSTM and CNN-Mamba architectures in photon dose calculations.
• Both BEV and patient-coordinate approaches achieved accurate dose calculations.
• BEV-based methods showed superior segment performance.
• CNN-ConvLSTM and CNN-Mamba offered comparable accuracy at reduced computational costs.