Radiotherapy dose distribution prediction for breast cancer using deformable image registration
Xue Bai1,2,3,4, Binbing Wang5,6,7, Shengye Wang5,6,7
1Key Laboratory of Radiation Physics and Technology, Ministry of Education, Institute of Nuclear Science and Technology, Sichuan University, Chengdu, 610064, China. baixue@zjcc.org.cn.
Biomedical Engineering Online
|May 31, 2020
Summary
This study developed and compared three methods for predicting 3D radiation dose distributions for breast cancer patients using historical treatment data. The weighted method (WEI_F) showed superior accuracy for organ-at-risk dose prediction, while the similar image (SIM) method excelled in spatial accuracy.
Area of Science:
- Medical Physics
- Radiation Oncology
- Image-guided Therapy
Background:
- Radiotherapy treatment planning relies on accurate dose prediction for quality assurance and automated planning.
- Dose-volume histograms (DVH) are standard for evaluating radiotherapy plans and prognosis, but 3D dose distributions offer more explicit insights into radiation effects.
Purpose of the Study:
- To predict the 3D dose distribution for breast cancer radiotherapy using deformable registration with historical treatment planning data.
- To evaluate and compare three distinct registration-based prediction approaches: SIM, WEI_A, and WEI_F.
Main Methods:
- A retrospective study involving an atlas cohort of 20 left-sided breast cancer patients treated with VMAT.
- Application of registration-based prediction techniques to 20 external patients.
- Quantification of dose prediction performance using metrics like ROI dose error, DSCs for prescription areas, and gamma metrics.
Main Results:
- The WEI_F method demonstrated superior dose prediction compared to WEI_A and outperformed SIM in organ-at-risk mean absolute difference (MAD).
- SIM showed better performance in Dice Similarity Coefficients (DSCs) and gamma metrics for predicting dose distributions.
- Specific MAD values for organs at risk (lung, heart) and plan target volume/spinal cord were reported for WEI_F and SIM methods.
Conclusions:
- The study successfully generated predicted dose distributions that closely approximated clinical treatment plans.
- The evaluated methods offer viable approaches for dose prediction in radiotherapy, with distinct strengths for different predictive aspects.
