Related Experiment Videos
A model for predicting lung cancer response to therapy.
Rebecca M Seibert1, Chester R Ramsey, J Wesley Hines
1Department of Radiation Oncology, Thompson Cancer Survival Center, Knoxville, TN 37916, USA. rseiber1@utk.edu
International Journal of Radiation Oncology, Biology, Physics
|January 24, 2007
Summary
A new locally weighted regression (LWR) model accurately predicts lung tumor volume changes during radiation therapy using early treatment data. This predictive model, utilizing serial megavoltage CT (MVCT) images, can help optimize cancer treatment by forecasting tumor response.
Area of Science:
- Radiation Oncology
- Medical Imaging
- Computational Biology
Background:
- Image-guided radiation therapy (IGRT) enables volumetric imaging during treatment.
- Predictive adaptive therapy aims to forecast tumor response for treatment optimization.
- Serial megavoltage CT (MVCT) provides data for monitoring tumor changes.
Purpose of the Study:
- Develop and validate a predictive model for lung tumor response during IGRT.
- Utilize serial MVCT data for early prediction of treatment outcomes.
- Assess the feasibility of adaptive therapy based on predicted tumor behavior.
Main Methods:
- A nonparametric, memory-based locally weighted regression (LWR) model was developed.
- Retrospective analysis of 480 serial MVCT images from 20 lung cancer lesions.
- Model prediction accuracy was tested using leave-one-out cross-validation.
Main Results:
- The LWR model predicted final tumor volume with an average error of 12%.
- Predictions were consistently within the 95% confidence interval.
- Optimal prediction days identified as 1, 2, 5, 9, 11, 12, 17, and 18.
Conclusions:
- The LWR model accurately predicted final tumor volume using early treatment data (8 days of observation).
- Quantified uncertainty in predictions supports potential for treatment optimization.
- This approach could enable more personalized and effective radiation therapy strategies.