Researchers led by Professor Fu Jin have developed an innovative approach that combines Monte Carlo simulations with deep learning to improve the speed and precision of quality assurance in radiation therapy. This method tackles a key issue in electronic portal imaging device (EPID)-based dose verification, where clinicians must balance rapid computations with reliable accuracy.
The Challenge in Radiation Therapy Verification
EPID serves as a vital tool for real-time, in vivo dose checks during treatments. Monte Carlo (MC) simulations stand as the benchmark for dose calculations, yet they present a trade-off: more simulated particles yield greater accuracy but demand extended processing times. Fewer particles speed up the process but introduce noise that undermines the results’ trustworthiness.
Integrating MC Simulations with Deep Learning
To overcome this, the team integrates the GPU-accelerated MC code ARCHER with the SUNet neural network, designed specifically for noise reduction. In tests using lung cancer intensity-modulated radiation therapy (IMRT) cases, they produced noisy EPID transmission dose data at varying particle counts: 1×106, 1×107, 1×108, and 1×109. SUNet trained on low-particle data, using the high-fidelity 1×109 dataset as the reference standard.
Impressive Performance Gains
The combined MC-deep learning (MC-DL) framework delivers superior speed and dosimetric precision. For the noisy 1×106-particle data, SUNet denoising elevates the structural similarity index (SSIM) from 0.61 to 0.95 and boosts the gamma passing rate (GPR) from 48.47% to 89.10%. The 1×107-particle data, an ideal speed-accuracy compromise, achieves an SSIM of 0.96 and GPR of 94.35% post-denoising, while 1×108 particles reach 99.55% GPR.
Denoising takes just 0.13–0.16 seconds, slashing total computation to 1.88 seconds for 1×107 particles and 8.76 seconds for 1×108. The resulting images show reduced graininess and smooth dose profiles that preserve essential clinical details, proving the technique’s effectiveness for radiotherapy QA.
Advancing Clinical Applications
This breakthrough proves especially valuable for online adaptive radiotherapy (ART), where quick dose checks reduce patient discomfort and account for anatomical shifts. The approach offers versatility: 1×107 particles suit urgent scenarios, while 1×108 ensures precision for complex cases.
“By integrating the accuracy of Monte Carlo simulation with the computational efficiency of deep learning, we have developed a practical solution that addresses the critical clinical need for rapid and reliable patient-specific quality assurance,” stated Professor Fu Jin. “This technology not only enhances existing radiation therapy workflows but also establishes a foundation for advanced applications, such as 3D dose reconstruction and broader implementation across diverse anatomical sites.”
Future efforts will extend the model to additional treatment sites, refine the SUNet architecture, and investigate other neural networks to enhance dose prediction.
Published in Nuclear Science and Techniques, DOI: https://doi.org/10.1007/s41365-026-01898-2
