PyDoseRT Photon: Physics-Guided Pencil-Beam Dose Calculation with Neural Priors and Residual Correction for CT and MRI
Abstract
We present a hybrid, physics-based analytical pencil-beam (PB) dose engine, augmented by two small frozen neural physics priors, followed by a 3-D convolutional residual-correction network (U-Net). We address the DoseRAD2026 (https://doserad2026.grand-challenge.org/) photon dose-prediction task with PyDoseRT Photon, a GPU PB engine implemented in PyTorch and corrected toward Monte Carlo (MC) accuracy in three learned stages of decreasing physical specificity. The engine reproduces the challenge's head-less MC source exactly where it can and models the patient with a beam-quality-indexed pencil kernel evaluated at the field's fluence-weighted radiological depth, and TERMA source scaling at the interaction site. Two tiny neural priors are trained through the frozen engine and then frozen themselves: a 39k-parameter 2-D fluence correction and a 48-parameter lateral heterogeneity correction mixing mass-conserving Gaussian redistribution operators. A compact 3-D U-Net (1.36M parameters) with a sequential refinement branch then predicts, per control point (CP) in the beam's-eye-view (BEV) frame, a bounded multiplicative gain and additive residual from seven channels. All learned stages are zero-initialized, so training starts from the analytical solution. For MRI, an nnU-Net regression model synthesizes a CT that enters the identical pipeline, with consecutively, the same trained corrector as the CT track. Design choices were driven by the challenge ranking, in which runtime carries double weight. The submitted method evaluated on our local CT and MR validation dataset achieved CP MAE 0.0086 and 0.0098, IDD distance 0.0011 and 0.0013, plan MAE 0.0024 and 0.0047, gamma pass rate (1%/1mm) 99.12% and 96.95%, with runtimes of 46s and 49s, respectively.
Related
- PyDoseRT Proton: A GPU Pencil-Beam Engine with a Convolutional Residual-Correction Network for Fast Proton Dose Calculation
- DoseRAD2026: A benchmark dataset for AI‐accelerated photon and proton dose calculation in radiotherapy
- Four-photon subtraction from picosecond squeezed vacuum demonstrating Wigner negativity without loss correction
- Microfacet-based physics-driven framework for single-photon LiDAR rough-surface scattering and detection
- Physics-anchored residual quantum neural networks for cloud-cover parameterization
BibTeX
@misc{simko2026pydosert,
title = {{PyDoseRT Photon: Physics-Guided Pencil-Beam Dose Calculation with Neural Priors and Residual Correction for CT and MRI}},
author = {Attila Simkó and Lukas Zimmermann and Hermann Fuchs and Gerd Heilemann},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.01085},
eprint = {2609.01085},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.01085},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - PyDoseRT Photon: Physics-Guided Pencil-Beam Dose Calculation with Neural Priors and Residual Correction for CT and MRI
AU - Attila Simkó
AU - Lukas Zimmermann
AU - Hermann Fuchs
AU - Gerd Heilemann
PY - 2026
DA - 2026/09/01
PB - arXiv
DO - 10.48550/arxiv.2609.01085
UR - https://arxiv.org/abs/2609.01085
AB - We present a hybrid, physics-based analytical pencil-beam (PB) dose engine, augmented by two small frozen neural physics priors, followed by a 3-D convolutional residual-correction network (U-Net). We address the DoseRAD2026 (https://doserad2026.grand-challenge.org/) photon dose-prediction task with PyDoseRT Photon, a GPU PB engine implemented in PyTorch and corrected toward Monte Carlo (MC) accuracy in three learned stages of decreasing physical specificity. The engine reproduces the challenge's head-less MC source exactly where it can and models the patient with a beam-quality-indexed pencil kernel evaluated at the field's fluence-weighted radiological depth, and TERMA source scaling at the interaction site. Two tiny neural priors are trained through the frozen engine and then frozen themselves: a 39k-parameter 2-D fluence correction and a 48-parameter lateral heterogeneity correction mixing mass-conserving Gaussian redistribution operators. A compact 3-D U-Net (1.36M parameters) with a sequential refinement branch then predicts, per control point (CP) in the beam's-eye-view (BEV) frame, a bounded multiplicative gain and additive residual from seven channels. All learned stages are zero-initialized, so training starts from the analytical solution. For MRI, an nnU-Net regression model synthesizes a CT that enters the identical pipeline, with consecutively, the same trained corrector as the CT track. Design choices were driven by the challenge ranking, in which runtime carries double weight. The submitted method evaluated on our local CT and MR validation dataset achieved CP MAE 0.0086 and 0.0098, IDD distance 0.0011 and 0.0013, plan MAE 0.0024 and 0.0047, gamma pass rate (1%/1mm) 99.12% and 96.95%, with runtimes of 46s and 49s, respectively.
ER -