Sharing a flexible urethral sparing SBRT prostate RapidPlan model and quantifying plan quality via dosimetric scorecard with clinical implementation

Jonathan Sackett, Anthony Magliari, Ryan Clark, Lesley Rosa, Kenny Guida, Riqiang Gao, Simon Arberet, Ali KamenRadiation 2025 · DOI

TL;DR

This paper introduces a novel Knowledge-Based Planning (KBP) model for prostate SBRT that uses a detailed dosimetric scorecard to ensure high-quality, consistent treatment plans, particularly focusing on sparing the urethra. This approach significantly improves plan quality compared to traditional methods and offers a flexible, single-click solution for clinical implementation.

What Problem Does This Paper Address?

Stereotactic Body Radiation Therapy (SBRT) for prostate cancer offers a potent treatment option by delivering high doses in fewer sessions. However, achieving optimal outcomes requires a delicate balance: maximizing radiation to the tumor while rigorously sparing critical organs like the rectum, bladder, and especially the urethra. The urethra is particularly sensitive, and damage can lead to significant urinary toxicity. Traditional manual treatment planning is time-consuming, prone to inter-planner variability, and struggles to consistently meet stringent clinical trial constraints. Existing KBP models, while improving efficiency, often lack the flexibility to adapt to diverse clinical scenarios and specific urethral sparing requirements. This paper aims to bridge this gap by developing a KBP model that is both highly effective and adaptable, guided by a comprehensive quality assessment tool.

Methods

The core of this research lies in the development of a KBP model named "ProstateSBRT-ARTIA-SIB40-36 Gy v1.1" using Varian's RapidPlan software. The model's development was guided by an iterative process, where initial plans were generated, scored using a custom dosimetric scorecard, and then used to retrain the model. This iterative refinement, driven by the scorecard's feedback, allowed for the progressive tuning of optimization objectives to achieve superior plan quality.

The key innovation is the 35-metric dosimetric scorecard. This scorecard, developed with input from multiple physicians, quantifies plan quality using a combination of Dose Volume Histogram (DVH) metrics. Each metric is assigned points based on piecewise linear scoring functions, culminating in a single score out of 229 points. This scorecard serves as an objective measure to guide the KBP model's development and evaluate its performance, ensuring it captures complex clinical intent.

The model was trained on 41 prostate SBRT cases and validated on 10 independent cases, testing both VMAT and IMRT delivery techniques.

Datasets

The study utilized a training dataset of 41 prostate SBRT cases from 6-arc Halcyon machines. These cases were carefully selected to represent a range of anatomical variations, prostate sizes, and treatment scenarios. Notably, half of the training set included cases with rectal spacers, while the other half did not, reflecting real-world clinical practice. The validation dataset consisted of 10 independent cases, also including scenarios with and without rectal spacers.

Key Insights and Ideas

  • Scorecard-Driven KBP is Powerful: The dosimetric scorecard isn't just an evaluation tool; it's an active driver of KBP model development. This iterative, scorecard-guided approach allows for the translation of complex clinical intent into actionable planning objectives, leading to demonstrably better plans.
  • Flexibility is Key for Clinical Adoption: A one-size-fits-all KBP model is insufficient. The developed model's ability to adapt to different dose levels, fractionation schemes, and even to perform minimal or no urethral sparing (as per clinical need) is crucial for its broad applicability.
  • "Single-Click" Planning is Achievable: The iterative refinement process, guided by the scorecard, aims to create a model that can generate high-quality plans with minimal user interaction, significantly reducing planning time and inter-planner variability.
  • Urethral Sparing is Quantifiable and Achievable: The scorecard specifically addresses urethral sparing, and the model's performance demonstrates that high-quality plans can be generated while meeting stringent urethral dose constraints.

Key Results

  • Significant Plan Quality Improvement: The KBP model demonstrated a substantial increase in average scorecard scores, rising from 166.2/229 for original clinical plans to 197.4/229 for KBP-generated plans.
  • Consistent High-Quality Plans: Validation cases showed consistent high-quality plan generation across both VMAT and IMRT techniques, indicating good generalizability.
  • Adaptability Demonstrated: The model proved flexible in adapting to various dose levels and clinical scenarios, including the ability to scale urethral dose constraints.
  • Clinical Validation: An early adoption test at an external institution showed that the model-generated plans outperformed manually optimized plans, achieving higher scores on an adjusted clinical scorecard.

Limitations and Future Work

  • Small Validation Set: The study acknowledges a limitation in the validation set size (10 patients, with only half having rectal spacers) and that validation was performed at a single external institution.
  • Iterative Bias: The iterative training process inherently creates a bias towards plans that optimize scorecard performance, which could make statistical significance testing of scorecard improvements potentially circular.
  • Future Directions: The authors propose expanding the model's capabilities to include boosting gross disease volumes and exploring direct optimization based on comprehensive dosimetric scorecards. Further work will also focus on quantifying the clinical benefit of this approach and exploring its use in large-scale synthetic data generation for AI training.

Summary updated

Abstract

The dosimetric scorecard-guided KBP model for prostate SBRT demonstrates the potential to generate high-quality treatment plans efficiently and consistently. This approach offers a powerful method for translating complex clinical intent into actionable planning objectives, potentially improving treatment quality and reducing inter-planner variability. Future work will focus on expanding the model's capabilities, including boosting gross disease volumes and exploring direct optimization based on comprehensive dosimetric scorecards.
BibTeX
@inproceedings{sackett2025sharing,
  author = {Jonathan Sackett and Anthony Magliari and Ryan Clark and Lesley Rosa and Kenny Guida and Riqiang Gao and Simon Arberet and Ali Kamen},
  title = {{Sharing a flexible urethral sparing SBRT prostate RapidPlan model and quantifying plan quality via dosimetric scorecard with clinical implementation}},
  booktitle = {Radiation},
  year = {2025},
  doi = {10.1186/s13014-025-02761-1},
}
Figure