Abstract Background Online adaptive radiation therapy necessitates independent dose verification systems for QA processes. Advanced irradiation techniques like IMRT, VMAT, and DWA require accurate Monte Carlo‐based dose calculations, where intra‐MLC transport plays a significant role due to complex beam‐shaping demands. Purpose This study develops an efficient ray‐tracing‐based intra‐MLC transport algorithm for independent dose verification in patient QA for online adaptive therapy. Method We developed a ray‐tracing‐based quasi‐analytical algorithm for particle transport in a multi‐leaf collimator (MLC) that considers photon attenuation and primary Compton scattering within the MLC. To achieve both efficiency and high accuracy in calculating the transparency within the MLC, the algorithm employs an intersection‐based ray‐tracing technique to calculate the path length of photons by determining intersections with MLC components on projection planes corresponding to the leaf motion and arrangement planes. The transmittance of photons is calculated based on the path length in the MLC and assigned as the weight of the photon, following Sieber's approach. To accelerate the method, two types of Russian roulette techniques were implemented. The first is a geometrical Russian roulette method based on the MLC aperture, which aims at reducing the computational burden of photons transported outside the irradiation field. The second applies the Russian roulette method to the generation of Compton photons, which suppresses the impact of large‐angle scattered Compton photons and further improves computational efficiency. We implemented the proposed transport calculation method for the MLC with Geant4 to calculate dose distributions in both digital phantoms and patient geometries. The proposed method was verified using the geometry of the Vero4DRT radiation therapy system (Hitachi High‐Tech Corporation., Tokyo, Japan). The fluence and dose distributions obtained from the method were compared with the results of full Geant4 simulations, where all stages of the calculation process were performed using Geant4. The computational efficiency of the MLC transport calculation was also analyzed to confirm its efficiency. Results The calculation results obtained using the proposed method reasonably reproduced the fluence and dose distributions calculated by full Geant4 simulations for both digital phantoms and patient geometries. The 95th percentile of the absolute point‐dose differences was within 2% for three‐dimensional voxels with dose exceeding 10% of the maximum dose obtained from the full Geant4 simulation, in the presence of propagated statistical uncertainties. For the MLC transport calculation efficiency, the introduction of the two Russian roulette techniques successfully suppressed the variation in computation time caused by field size and achieved a speed‐up of 56–160 times compared to Geant4. Conclusion In this study, we proposed an efficient quasi‐analytical MLC transport calculation method using an intersection‐based ray‐tracing technique. The developed code was integrated with Geant4, and comparisons with full Geant4 simulations confirmed that the calculation accuracy was sufficient. The algorithm can be integrated with GPU‐accelerated Monte Carlo‐based dose calculation codes designed for voxel geometries, which have been actively developed in recent years, making it a useful tool for independent dose verification in online adaptive radiotherapy.
Hirayama et al. (Thu,) studied this question.