Predicting objective function weights from patient anatomy in prostate IMRT treatment planning

Taewoo Lee, Muhannad Hammad, Timothy C. Y. Chan, Tim Craig, Michael B. SharpeMedical 2013 · DOI

Abstract

This study demonstrates a proof of concept that patient anatomy can be used to predict appropriate objective function weights for treatment planning. In the long term, such geometry-driven weights may serve as a starting point for iterative treatment plan design or may provide information about the most clinically relevant region of the Pareto surface to explore.

Introduction

Intensity-Modulated Radiation Therapy (IMRT) planning is a complex balancing act. Planners must deliver a therapeutic dose to the target volume while minimizing exposure to surrounding Organs at Risk (OARs). This multi-criteria optimization (MCO) problem is typically solved by assigning weights to various objective functions. Currently, these weights are determined through a manual, iterative "trial-and-error" process, which is time-consuming and prone to inter-planner variability. This research explores whether patient-specific anatomy can be used to predict these "ideal" weights directly, automating a significant portion of the planning process.

Methodology

The researchers developed a predictive pipeline consisting of three main components:

  1. Inverse Optimization (IOM): To establish a ground truth, the team applied IOM to 24 high-quality historical prostate cancer plans. This process reverse-engineers the finished dose distributions to identify the exact objective function weights ($\alpha$) that would produce such a plan.
  2. Anatomical Quantification (OVH): The Overlap Volume Histogram (OVH) was used to characterize the spatial relationship between the target (prostate) and OARs (bladder and rectum). The study investigated various expansion distances to find the most predictive geometric metric.
  3. Statistical Modeling: A logistic regression model was used to map the geometric features to the rectum weight. The femoral heads were assigned a fixed low weight (1%), focusing the model on the primary trade-off between the bladder and rectum.
Comparison of rectum weight correlation across different anatomical expansion distances (0 cm to 2 cm).

Data and Validation

The study utilized 24 prostate IMRT plans from the Princess Margaret Cancer Centre. Validation was performed using a leave-one-out cross-validation strategy. The performance of the predicted weights was benchmarked against both the "ideal" IOM weights and a population-average weight vector.

Key Insights

A critical finding was that the ratio of rectum-to-bladder overlap with the target at a 1 cm expansion ($OV_1$) served as a highly effective predictor for the rectum weight. This metric captures the "difficulty" of sparing one organ relative to the other.

Detailed view of the correlation between rectum weight and the $OV_1$ metric (1 cm expansion), which was identified as the most powerful predictor.

The high correlation suggests that clinical intent is often a consistent response to specific anatomical constraints. By formalizing this relationship, the model standardizes the planning logic, ensuring that patients with similar geometries receive consistent treatment strategies regardless of the individual planner.

Results

The predictive model demonstrated significant improvements over traditional methods:

  • Weight Accuracy: Predicted weight vectors were six times closer to the ideal IOM weights compared to using a population average.
  • Dosimetric Performance: Plans generated using predicted weights achieved dose metrics (e.g., V54.3 Gy and V70.0 Gy for OARs) within 5 percentage points of the original clinical plans.
  • Statistical Significance: A one-sided sign test confirmed that anatomy-based weights significantly outperformed average weights, validating the need for personalized planning.

Limitations and Future Work

While successful as a proof-of-concept for prostate cases, the model currently only handles a bi-criteria trade-off. Future work will need to address more complex anatomical sites, such as the head and neck, which involve numerous OARs. Additionally, expanding the dataset beyond 24 patients will improve the model's robustness to rare anatomical variations. Ultimately, these predicted weights could provide an "informed starting point" for navigating the Pareto surface in advanced radiation therapy planning.

Summary updated

Abstract

This study demonstrates a proof of concept that patient anatomy can be used to predict appropriate objective function weights for treatment planning. In the long term, such geometry-driven weights may serve as a starting point for iterative treatment plan design or may provide information about the most clinically relevant region of the Pareto surface to explore.
BibTeX
@inproceedings{lee2013predicting,
  author = {Taewoo Lee and Muhannad Hammad and Timothy C. Y. Chan and Tim Craig and Michael B. Sharpe},
  title = {{Predicting objective function weights from patient anatomy in prostate IMRT treatment planning}},
  booktitle = {Medical},
  year = {2013},
  doi = {10.1118/1.4828841},
}

Loading related content...

Figure