From Radiotherapy Dataset to Cherenkov Dose Map: A Digital Linac Monte Carlo Simulation Pipeline

Abstract

Background: Cherenkov-based radiotherapy monitoring depends on the relationship between dose and locally generated Cherenkov photons, before optical transport and detection. Purpose: To develop a DICOM RT-driven Geant4 digital linac Monte Carlo framework that scores dose and Cherenkov creation together and to evaluate their spatial correspondence. Methods: The framework converts DICOM CT, structure, plan, and dose objects into beam-delivery states and scores energy deposition and Cherenkov photon creation on a common 2.5-mm isotropic grid in a water-equivalent phantom. Static fields, different beam qualities, and a two-arc VMAT plan were tested. Monte Carlo dose was compared with TPS RTDOSE from Acuros XB and AAA; dose-Cherenkov associations and spatial agreement were assessed at voxel, plane, and profile levels. Results: Full-volume 3D dose gamma passing rates (3%/3 mm, 10% threshold) were 97.05%-98.86% for Acuros XB-referenced field-size and beam-quality cases (97.47% with AAA) and 97.84% for VMAT. Cherenkov creation correlated strongly with dose, with voxel $r^2 \geq 0.9890$ and projection/profile $r^2 \geq 0.9942$ in all cases. Dose-Cherenkov 3D gamma passing rates were 97.97%-99.16% for $10\times10$ 6X, 10X, and 15X fields and 96.59% for VMAT, but fell to 71.22% for $5\times5$ fields and 32.82% for 6FFF. Deviations were most apparent near small-field edges and along depth for 6FFF. Conclusions: In a homogeneous water phantom, local Cherenkov creation broadly tracks local dose but shows systematic differences at beam edges and with FFF beam quality. This framework supports future studies of optical transport and surface-based Cherenkov dosimetry.