3056 - Efficient High-Quality IMRT Dose Generation Using a Multi-Head Attention Enhanced Recurrent Mamba Network

12:30pm - 01:45pm ET

Poster Hall - Exhibit Hall A

Screen: 29

POSTER

Presenter(s)

Nan Li, PhD - Student, BeiJing, BeiJing

N. Li¹, Y. Liu², T. Lv³, G. Zhang⁴, and S. Xu²; ¹Bei Hang university China, BeiJing, China, ²National Cancer Center/ National Clinical Research Center for Cancer/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College, Beijing, China, ³Department of Radiotherapy, Beijing Hospital, Beijing, China, ⁴School of Physics, Beihang University, Beijing, China

Purpose/Objective(s):

Monte Carlo (MC) simulation serves as the gold standard for dose calculation in radiation therapy, as it can accurately reproduce the interaction processes between photons and human tissues in intensity-modulated radiation therapy (IMRT) and generate high-precision dose distributions. However, it is limited by low computational efficiency and long computation time. In clinical radiotherapy planning, the conventional default dose calculation grid is 3 mm. If a finer grid (e.g., 1 mm) is used to improve simulation accuracy, the computation time increases exponentially. Deep learning models provide an effective solution to this technical contradiction. To satisfy the dual requirements of high precision and high efficiency in dose simulation, this study aims to develop and train a deep learning-based dose calculation model for high-precision, high-efficiency dose distribution generation.

Materials/Methods:

This study proposes a high-precision dose generation strategy based on a multi-head attention-enhanced recurrent Mamba network. The proposed method seamlessly integrates the gamma pass rate loss function into the network training framework, enabling efficient and accurate conversion of dose simulation results from a 3 mm grid (meeting basic clinical requirements) into high-resolution, high-precision dose distributions.

Results:

The proposed method can super-resolve the dose distribution from a 3 mm grid to 1 mm grid precision within 30 seconds. Validated using gamma analysis with a 10% dose threshold, the gamma pass rates achieved by this method were 99.11%±0.71%, 97.52%±0.61%, and 96.31%±1.03% under the criteria of 3 mm/3%, 2 mm/2%, and 1 mm/1%, respectively.

Conclusion:

Our method resolves the accuracy–efficiency dilemma in MC-based IMRT dose calculation. It achieves 3 mm to 1 mm dose super-resolution within 30 s with gamma pass rates above 95%, providing an efficient clinical solution for high-precision dose calculation with strong application potential.