A proof-of-concept implementation of a patient-specific Geant4 Monte Carlo dose evaluation framework for SFRT.

Abstract

Accurate dose calculation in spatially fractionated radiotherapy (SFRT) is challenging 
because highly heterogeneous dose distributions and charged-particle disequilibrium can 
reduce the accuracy of fast analytical algorithms, including the anisotropic analytical 
algorithm (AAA) used in commercial treatment planning systems (TPS). Monte Carlo 
(MC) methods are considered the reference standard for complex dose calculations, but 
patient-specific end-to-end studies of SFRT under realistic clinical conditions remain scarce. 
In this proof-of-concept study, we implemented a patient-specific Geant4-based MC 
framework for SFRT dose calculation and applied it to one clinical treatment plan. Patient 
geometry was reconstructed from planning CT DICOM data. A Varian-provided 6 MV 
flattening-filter-free (FFF) phase-space source upstream of the secondary collimators was 
used to simulate the TrueBeam linear accelerator. Secondary collimators, including the jaws 
and a 120-leaf multileaf collimator (MLC), were modeled explicitly. Clinical volumetric 
modulated arc therapy (VMAT) control points were sequentially accumulated to enable 
end-to-end simulation. The three-dimensional MC dose distribution was quantitatively 
compared with the TPS-AAA calculation. Open-field water-phantom measurements yielded 
percentage-depth-dose errors within 3% and dose-profile errors within 4% for three standard 
square fields. No measurements of MLC-shaped fields or patient-specific measurement-based 
quality assurance (QA) were performed. For the investigated SFRT plan, three-dimensional 
global gamma passing rates using MC as the reference were 94.62%, 97.82%, and 99.40% 
under the 2%/2 mm, 3%/2 mm, and 3%/3 mm criteria, respectively, with a 10% dose 
threshold. The PVDR values were 2.050-2.620 for AAA and 1.884-2.454 for Geant4. 
Dose-volume histogram (DVH) analysis showed slight underdosage in the gross tumor 
volume (GTV), while profile comparisons showed case-specific agreement between AAA and 
MC. These results demonstrate the technical feasibility of the framework and provide a 
transparent methodological basis for future validation studies. The framework is not 
intended to replace routine TPS calculations or serve as a commissioned clinical QA system.