Related Experiment Video
Updated: Jan 6, 2026
08:17
Dynamic Lung Tumor Tracking for Stereotactic Ablative Body Radiation Therapy
Published on: June 7, 2015
16.1K
Prediction of lung tumor motion using nonlinear autoregressive model with exogenous input.
Kai Jiang1,2, Fumitake Fujii1,3, Takehiro Shiinoki4
1Graduate School of Science and Technology for Innovation, Yamaguchi University, 2-16-1, Tokiwa-dai, Ube 755-8611, Japan.
Physics in Medicine and Biology
|October 2, 2019
Summary
This study developed a lung tumor position predictor using a nonlinear autoregressive model with exogenous input (NARX) for dynamic tumor tracking radiotherapy. The NARX predictor accurately forecasts tumor movement, compensating for linear accelerator control lag.
Area of Science:
- Medical Physics
- Radiotherapy Technology
- Computational Biology
Background:
- Dynamic tumor tracking radiotherapy (DTT-RT) requires precise lung tumor localization.
- Commercial linear accelerators have a 50-500 ms positioning lag in multi-leaf collimator (MLC) control.
- Accurate prediction of future tumor positions is crucial for effective DTT-RT.
Purpose of the Study:
- To develop and evaluate a lung tumor position predictor for DTT-RT.
- To compensate for MLC positioning lag using predictive modeling.
- To assess the feasibility of generating future gating cubes for DTT-RT.
Main Methods:
- A nonlinear autoregressive model with exogenous input (NARX) was employed.
- Patient-specific NARX models were trained using lung tumor motion trajectories.
- Three prediction horizons (600 ms, 800 ms, 1 s) were investigated.
- Performance metrics included RMS prediction errors and trajectory coverage rates.
Main Results:
- The NARX predictor demonstrated effective lung tumor motion prediction.
- Rates of coverage for the entire tumor trajectory were high across prediction horizons.
- Mean coverage rates were [Formula: see text]%, [Formula: see text]%, and [Formula: see text]% for 600 ms, 800 ms, and 1 s horizons, respectively.
Conclusions:
- The proposed NARX predictor is a viable tool for lung tumor position prediction in DTT-RT.
- The predictor effectively compensates for MLC positioning lag.
- This approach enhances the feasibility of implementing DTT-RT by enabling accurate gating cube generation.
