Differentiable Autoplanning Algorithm for Proton Therapy Based on Deep Learning: A Proof-of-Concept
- 1Department of Physics, Sapienza University of Rome, Rome, Italy
- 2Istituto Superiore di Sanità, Rome, Italy
Abstract
Purpose
This study presents the proof-of-concept of a novel framework for fully automated proton therapy treatment planning that enables the simultaneous optimization of all major beam parameters, including fluences, energies, and field orientations, within a single differentiable programming approach. The objective is to overcome current limitations of conventional treatment planning systems, which restrict optimization to fluence alone.Materials and methods
A differentiable deep-learning (DL) dose engine based on Graph Neural Networks was employed to emulate Monte Carlo (MC) dose deposition of proton pencil beams with millisecond inference time and high accuracy. The model predicts energy deposition within a cylindrical scoring volume around each beam, which is subsequently mapped to the CT reference frame using differentiable geometric transformations and interpolation. Exploiting end-to-end differentiability, exact gradients of dose with respect to beam energies, orientations, and fluences were computed and used for plan optimization. A hybrid strategy combining Covariance Matrix Adaptation Evolution Strategy (CMA) for global exploration and the Adam gradient-based optimizer for local refinement was implemented. Proof-of-concept evaluations were conducted in homogeneous water phantom geometries and in a head-and-neck phantom CT featuring single- and dual-field configurations targeting a defined planning target volume while sparing organs at risk.Results
The DL dose engine reproduced high-statistics MC simulated dose distributions with mean relative errors of the order of percent, with a computational efficiency compatible with iterative optimization. The joint optimization procedure reliably converged from random initial conditions to reproducible solutions in both single- and dual-field setups, generating clinically meaningful Spread-Out Bragg peaks with homogeneous planning target volume coverage and effective organs at risk sparing.Conclusion
This work demonstrates the feasibility of fully joint, gradient-based optimization of all key proton therapy beam parameters using a differentiable DL dose engine. The proposed framework provides a foundation for next-generation automated treatment planning systems and motivates future extension to patient-specific, clinically realistic scenarios.
Related
- Loco-regional radiopharmaceutical therapy in breast carcinoma using [177Lu]Lu-microseed implants: Development to proof-of-concept
- Using Routine Ventilator Data to Visualize and Quantify Dynamic Compliance Trends Post-Surfactant Therapy in Preterm Infants: A Proof-of-Concept Study
- From endovascular therapy to perivascular therapy: A new concept and perspective
- A differentiable multi-physics learning model for robust single-image dehazing under complex atmospheric conditions
- The content of biomedical papers predicts future clinical trials: a proof-of-concept framework tested on CAR-T cell therapy
BibTeX
@article{arsini2026differentiable,
title = {{Differentiable Autoplanning Algorithm for Proton Therapy Based on Deep Learning: A Proof-of-Concept}},
author = {Lorenzo Arsini and Barbara Caccia and Andrea Ciardiello and Riccardo Faccini and Stefano Giagu and Francesca Nicolanti and Carlo Mancini-Terracciano},
journal = {International journal of particle therapy},
year = {2026},
month = oct,
doi = {10.1016/j.ijpt.2026.101967},
url = {https://europepmc.org/article/PMC/PMC13718767},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Differentiable Autoplanning Algorithm for Proton Therapy Based on Deep Learning: A Proof-of-Concept
AU - Lorenzo Arsini
AU - Barbara Caccia
AU - Andrea Ciardiello
AU - Riccardo Faccini
AU - Stefano Giagu
AU - Francesca Nicolanti
AU - Carlo Mancini-Terracciano
PY - 2026
DA - 2026/10/09
JO - International journal of particle therapy
DO - 10.1016/j.ijpt.2026.101967
UR - https://europepmc.org/article/PMC/PMC13718767
AB - Purpose This study presents the proof-of-concept of a novel framework for fully automated proton therapy treatment planning that enables the simultaneous optimization of all major beam parameters, including fluences, energies, and field orientations, within a single differentiable programming approach. The objective is to overcome current limitations of conventional treatment planning systems, which restrict optimization to fluence alone.Materials and methods A differentiable deep-learning (DL) dose engine based on Graph Neural Networks was employed to emulate Monte Carlo (MC) dose deposition of proton pencil beams with millisecond inference time and high accuracy. The model predicts energy deposition within a cylindrical scoring volume around each beam, which is subsequently mapped to the CT reference frame using differentiable geometric transformations and interpolation. Exploiting end-to-end differentiability, exact gradients of dose with respect to beam energies, orientations, and fluences were computed and used for plan optimization. A hybrid strategy combining Covariance Matrix Adaptation Evolution Strategy (CMA) for global exploration and the Adam gradient-based optimizer for local refinement was implemented. Proof-of-concept evaluations were conducted in homogeneous water phantom geometries and in a head-and-neck phantom CT featuring single- and dual-field configurations targeting a defined planning target volume while sparing organs at risk.Results The DL dose engine reproduced high-statistics MC simulated dose distributions with mean relative errors of the order of percent, with a computational efficiency compatible with iterative optimization. The joint optimization procedure reliably converged from random initial conditions to reproducible solutions in both single- and dual-field setups, generating clinically meaningful Spread-Out Bragg peaks with homogeneous planning target volume coverage and effective organs at risk sparing.Conclusion This work demonstrates the feasibility of fully joint, gradient-based optimization of all key proton therapy beam parameters using a differentiable DL dose engine. The proposed framework provides a foundation for next-generation automated treatment planning systems and motivates future extension to patient-specific, clinically realistic scenarios.
SN - 2331-5180
ER -