Dynamic control of carbon-ion radiation dose distribution amid respiratory motion challenges: A review of prediction research using geant4 and deep learning physics-constrained models.
- 1School of Medicine, Faculty of Health and Medical Sciences, Taylor's University, Subang Jaya, Malaysia.
- 2School of Computing & Technology, Asia Pacific University of Technology & Innovation, Kuala Lumpur, Malaysia.
Abstract
Background
Carbon ion radiotherapy (CIRT) offers superior dose conformity and biological effectiveness; however, its precision is highly sensitive to respiratory motion, leading to dose uncertainty and interplay effects that compromise treatment accuracy.Methods
A comprehensive literature synthesis was conducted on studies employing motion-resolved imaging techniques (e.g. SMEIR), dynamic dose prediction, 4D robust planning, and hybrid adaptive methods for CIRT. The study emphasised the use of Geant4 Monte Carlo simulations combined with deep learning models, including physics-informed neural networks, to simulate and predict dose perturbations caused by respiratory motion. This review also examined clinical treatment planning systems, radiobiological modelling, and adaptive strategies to mitigate the effects of motion.Results
The integration of Geant4 with deep learning enabled high-fidelity, computationally efficient dose prediction under respiratory motion, improving accuracy in dose localization and adaptation. Motion management techniques such as respiratory gating, rescanning, and 4D robust optimization demonstrated enhanced tumor control and normal tissue sparing. Hybrid models combining Monte Carlo data and AI inference show promise for near-real-time adaptive therapy. Despite these advances, challenges remain, including limited motion-rich datasets, computational demands for real-time recalculations, and the complexity of incorporating variable relative biological effectiveness (RBE) into adaptive planning.Conclusions
Combining Geant4-based Monte Carlo simulations with deep learning frameworks shows emerging promise for controlling carbon ion dose distributions amid respiratory motion, though direct clinical validation for this combination remains limited. These hybrid models may enhance the precision and adaptability of CIRT, offering improved outcomes. Future efforts should focus on clinical validation, standardization, and overcoming computational and data limitations to facilitate clinical translation of motion-adaptive CIRT. These approaches remain at Technology Readiness Levels (TRL 3-5) and require prospective clinical validation before routine implementation.
Related
- Surface-guided Radiation Therapy in Clinical Practice: A Review with Perspectives on Implementation in Resource-constrained Settings
- From Link Availability to Deadline-Constrained Key Service: A Feasibility Framework for Mobile Free-Space Quantum Key Distribution
- Enhancing Extubation Failure Prediction with LLM-Derived Features from Respiratory Therapy Clinical Notes
- Prediction of biological radiation effects based on ionization clusters (nanodosimetry)
- Physics-Constrained Neural Surrogate for Domain Growth Prediction in Systems with Conserved Kinetics
BibTeX
@article{ding2026dynamic,
title = {{Dynamic control of carbon-ion radiation dose distribution amid respiratory motion challenges: A review of prediction research using geant4 and deep learning physics-constrained models.}},
author = {Jianhua Ding and Peng Li and Chai Hong Yeong},
journal = {Journal of applied clinical medical physics},
year = {2026},
month = oct,
doi = {10.1002/acm2.70766},
url = {https://europepmc.org/article/MED/42900519},
note = {CC BY},
}
RIS
TY - JOUR
TI - Dynamic control of carbon-ion radiation dose distribution amid respiratory motion challenges: A review of prediction research using geant4 and deep learning physics-constrained models.
AU - Jianhua Ding
AU - Peng Li
AU - Chai Hong Yeong
PY - 2026
DA - 2026/10/30
JO - Journal of applied clinical medical physics
DO - 10.1002/acm2.70766
UR - https://europepmc.org/article/MED/42900519
AB - Background Carbon ion radiotherapy (CIRT) offers superior dose conformity and biological effectiveness; however, its precision is highly sensitive to respiratory motion, leading to dose uncertainty and interplay effects that compromise treatment accuracy.Methods A comprehensive literature synthesis was conducted on studies employing motion-resolved imaging techniques (e.g. SMEIR), dynamic dose prediction, 4D robust planning, and hybrid adaptive methods for CIRT. The study emphasised the use of Geant4 Monte Carlo simulations combined with deep learning models, including physics-informed neural networks, to simulate and predict dose perturbations caused by respiratory motion. This review also examined clinical treatment planning systems, radiobiological modelling, and adaptive strategies to mitigate the effects of motion.Results The integration of Geant4 with deep learning enabled high-fidelity, computationally efficient dose prediction under respiratory motion, improving accuracy in dose localization and adaptation. Motion management techniques such as respiratory gating, rescanning, and 4D robust optimization demonstrated enhanced tumor control and normal tissue sparing. Hybrid models combining Monte Carlo data and AI inference show promise for near-real-time adaptive therapy. Despite these advances, challenges remain, including limited motion-rich datasets, computational demands for real-time recalculations, and the complexity of incorporating variable relative biological effectiveness (RBE) into adaptive planning.Conclusions Combining Geant4-based Monte Carlo simulations with deep learning frameworks shows emerging promise for controlling carbon ion dose distributions amid respiratory motion, though direct clinical validation for this combination remains limited. These hybrid models may enhance the precision and adaptability of CIRT, offering improved outcomes. Future efforts should focus on clinical validation, standardization, and overcoming computational and data limitations to facilitate clinical translation of motion-adaptive CIRT. These approaches remain at Technology Readiness Levels (TRL 3-5) and require prospective clinical validation before routine implementation.
SN - 1526-9914
ER -