Early generation dynamic and static proton arc treatment planning algorithms assessment in oropharyngeal cancer patients

Bas A. de Jong, Lewei Zhao, Peilin Liu, Viktor Wase, Otte Marthin, Johan Sundström, Gang Liu, Rohan Deroniyagala, Xiaoqiang Li, Xiaoda Cong, Guillaume Janssens, Erik Engwall, Erik Weyer, Xuanfeng Ding, Johannes A. Langendijk, Stefan BothMedical 2025 · DOI

TL;DR

This paper evaluates early-generation proton arc therapy (PAT) planning algorithms for oropharyngeal cancer, finding they significantly reduce toxicity and treatment time compared to standard intensity-modulated proton therapy (IMPT) and volumetric modulated arc therapy (VMAT). While PAT shows promise, its robustness to inter-fraction anatomical changes is a key area for improvement, with range shifters offering a potential solution.

What Problem Does This Paper Address?

Oropharyngeal cancer (OPC) treatment often involves radiation therapy, aiming to deliver a high dose to the tumor while sparing surrounding healthy tissues to minimize side effects and preserve quality of life. Current standard treatments include volumetric modulated arc therapy (VMAT) and intensity-modulated proton therapy (IMPT). Proton arc therapy (PAT) is an emerging technique that utilizes a larger number of gantry angles, potentially offering further improvements in dose distribution and toxicity reduction, especially for complex target shapes common in OPC. However, concerns exist regarding PAT's robustness to patient movement and anatomical changes during treatment, which can impact target coverage. This paper addresses the need to assess and compare the performance of different early-generation PAT planning algorithms (static and dynamic) against established IMPT and VMAT techniques, focusing on toxicity, delivery time, and inter-fraction robustness.

Methods

The study evaluated three PAT planning strategies:

  • Energy Layer Filtration (ELF): A static PAT approach that iteratively reduces energy layers in an IMPT plan.
  • Spot scanning Proton Arc (SPArc): A dynamic PAT approach.
  • Early Layer and Spot Assignment (ELSA): Another dynamic PAT approach.

These algorithms were used to generate PAT plans for six OPC patients who had previously received IMPT. The study also generated VMAT and IMPT plans for comparison. All proton plans were robustly optimized to account for setup and density uncertainties. Key metrics assessed included:

  • Expected Plan Toxicity: Evaluated using Normal Tissue Complication Probability (NTCP) models for dysphagia and xerostomia.
  • Delivery Time: Modeled using experimental machine log-files and gantry dynamics.
  • Inter-fraction Robustness: Assessed by evaluating target coverage on weekly repeated CT scans, both fraction-wise and cumulatively over the treatment course.

Datasets

The study used data from six OPC patients who had previously undergone IMPT treatment. No new datasets were introduced; rather, existing patient data was used to generate and compare different treatment plans.

Key Insights and Ideas

  • PAT's Toxicity Advantage: The study strongly suggests that PAT, across all tested algorithms, offers a significant reduction in NTCP for dysphagia and xerostomia compared to both IMPT and VMAT. This is a crucial finding, as reducing these toxicities directly translates to improved patient quality of life.
  • Dynamic PAT is Faster: Dynamic PAT delivery (SPArc and ELSA) was found to be faster than static PAT (ELF), and both were generally faster than clinical IMPT delivery. This efficiency gain is important for clinical workflow and patient throughput.
  • Robustness is a Challenge: A significant finding is that PAT plans, while offering dose benefits, are more susceptible to inter-fraction anatomical and setup changes than IMPT plans. This "fraction-wise robustness" is a critical consideration for clinical implementation.
  • Range Shifters as a Solution: The addition of anterior oblique fields with range shifters in the ELF static PAT plans demonstrated a marked improvement in both fraction- and course-wise target coverage, suggesting a practical method to enhance PAT robustness.
  • Algorithm Nuances: While all PAT algorithms showed benefits, there were subtle differences. SPArc and ELF showed slightly larger toxicity reductions than ELSA. The delivery time varied, with ELSA being the fastest dynamic option.

Key Results

  • NTCP Reduction: PAT plans showed significant (p < 0.05) reductions in NTCP for dysphagia and xerostomia grade ≥ 2 compared to IMPT and VMAT. For xerostomia, reductions ranged from 8.7% to 9.8% for 360-layer PAT plans compared to VMAT. For dysphagia, reductions were even higher, up to 13.1%.
  • Delivery Time: Dynamic PAT delivery times averaged around 7.6-8.2 minutes, significantly shorter than clinical IMPT delivery times (average 11.4 ± 2.1 minutes). Static PAT delivery times were slightly longer but still generally shorter than IMPT.
  • Target Coverage Robustness: Fraction-wise target coverage was worse in PAT plans compared to IMPT. However, course-wise coverage was less impacted. The addition of anterior oblique fields with range shifters improved target coverage for ELF plans.
  • Optimization Time: PAT plan optimization took more than twice as long as IMPT and VMAT, with SPArc being the most time-consuming, though this was partly attributed to its optimization setup.

Limitations and Future Work

  • Inter-fraction Robustness: The paper acknowledges that PAT's reduced inter-fraction robustness is a significant challenge. While range shifters helped improve robustness for static PAT, further investigation into systematic patient positioning and anatomical changes is needed, potentially requiring adaptive replanning strategies.
  • Algorithm Development: The study focused on "early generation" PAT algorithms. Future work should explore more advanced algorithms and their impact on robustness and efficiency.
  • Generalizability: The findings are specific to OPC patients. Further studies are needed to assess PAT robustness for other treatment sites.
  • Optimization Time Comparison: The direct comparison of SPArc optimization time was limited by its implementation and hardware, suggesting that proper integration into newer TPS could significantly reduce this time.
  • Future Directions: The authors suggest exploring further reductions in energy layers for dynamic PAT to potentially decrease delivery time without significantly increasing NTCP. They also highlight the need for further investigation into the relationship between energy layers, energy up-switches, and robustness in dynamic PAT. The potential for online adaptive replanning to ensure target coverage in PAT is also mentioned as a promising avenue.

Summary updated

Abstract

The tested static and dynamic PAT planning strategies showed similar significant reductions in NTCP compared to IMPT and VMAT. Estimated PAT delivery times were shorter compared to times for current delivery procedures of clinical IMPT plans. Dynamic PAT delivery is faster than static PAT. Fraction-wise robustness suffered more than course-wise robustness due to anatomical changes found in repeated CTs. Range shifters can be employed to improve PAT plan robustness.
BibTeX
@inproceedings{jong2025early,
  author = {Bas A. de Jong and Lewei Zhao and Peilin Liu and Viktor Wase and Otte Marthin and Johan Sundström and Gang Liu and Rohan Deroniyagala and Xiaoqiang Li and Xiaoda Cong and Guillaume Janssens and Erik Engwall and Erik Weyer and Xuanfeng Ding and Johannes A. Langendijk and Stefan Both},
  title = {{Early generation dynamic and static proton arc treatment planning algorithms assessment in oropharyngeal cancer patients}},
  booktitle = {Medical},
  year = {2025},
  doi = {10.1002/mp.17916},
}
Figure