Related Experiment Video
Updated: May 16, 2026
08:54
Treatment of Liver Metastases Using an Internal Target Volume Method for Stereotactic Body Radiotherapy
Published on: May 8, 2018
Predicting dose-volume histograms for organs-at-risk in IMRT planning
Lindsey M Appenzoller1, Jeff M Michalski, Wade L Thorstad
1Department of Radiation Oncology, Washington University, St. Louis, MO, USA.
Medical Physics
|December 13, 2012
Summary
This study developed a mathematical tool to predict organ-at-risk (OAR) dose-volume histograms (DVHs) in intensity modulated radiotherapy (IMRT) planning. The models accurately identify suboptimal plans, enabling improved OAR sparing and treatment quality.
Area of Science:
- Medical Physics
- Radiation Oncology
- Computational Biology
Background:
- Intensity modulated radiotherapy (IMRT) planning involves complex optimization to balance tumor coverage and organ-at-risk (OAR) sparing.
- Variability in IMRT plan quality can arise from differences in planning techniques and anatomical interpretations.
- Predictive models are needed to establish achievable OAR dose constraints and ensure consistent treatment plan quality.
Purpose of the Study:
- To develop a quality control (QC) tool using mathematical models to predict achievable OAR dose-volume histograms (DVHs).
- To reduce variability and improve treatment plan quality in intensity modulated radiotherapy (IMRT).
- To leverage individual patient anatomy for accurate DVH prediction.
Main Methods:
- A mathematical framework was derived correlating dose to minimum distance from a voxel to the planning target volume (PTV) surface.
- Skew-normal probability distributions were used to model subvolume dose distributions, with DVH prediction models developed via polynomial fitting.
- Organ-specific average models were trained using IMRT plans for prostate and head-and-neck cancers, with refinement based on outlier detection using restricted sums of residuals (RSR).
Main Results:
- Average models demonstrated good agreement with training and validation cohorts for rectum, bladder, and parotid DVHs (mean SR near zero).
- Refined models showed strong correlations between predicted and realized OAR sparing gains after replanning (r=0.92 for rectum, r=0.88 for bladder, r=0.84 for parotids).
- The models successfully identified suboptimal plans, indicating potential for dosimetric improvements.
Conclusions:
- Mathematical models can accurately predict achievable OAR DVHs based on patient anatomy using modest training cohorts.
- The developed QC tool effectively identifies plans with potential for improved OAR sparing.
- Clinical implementation is underway to assess the impact of this technique on real-time IMRT quality control.
