Prediction of respiratory motion with wavelet-based multiscale autoregression

Area of Science:

• Medical physics
• Robotics
• Signal processing

Background:

• Robotic radiosurgery utilizes a robot-guided photon beam for tumor ablation.
• Respiratory motion introduces inaccuracies in treatment delivery due to system delays.
• Accurate motion prediction is crucial for enhancing treatment precision.

Purpose of the Study:

• To develop and evaluate an advanced prediction algorithm for respiratory motion in robotic radiosurgery.
• To improve the accuracy of tumor ablation by compensating for system delays caused by patient movement.
• To enhance the performance of motion prediction models in real-time treatment scenarios.

Main Methods:

• A wavelet-based multiscale autoregressive prediction method was employed.
• The algorithm was enhanced with an exponential averaging parameter for improved signal dependency handling.
• The Moore-Penrose pseudo-inverse was utilized to address system matrix irregularities.
• Performance was evaluated against normalized Least Mean Squares (LMS) predictors.

Main Results:

• The novel algorithm demonstrated significant improvements in test cases, outperforming normalized LMS predictors by up to 50%.
• Real patient data showed an average improvement of 5% to 10% in treatment accuracy.
• The enhanced prediction method effectively compensated for system delays and motion variations.

Conclusions:

• The proposed wavelet-based prediction method offers a substantial advancement for robotic radiosurgery.
• Improved motion prediction leads to more accurate tumor targeting and potentially better patient outcomes.
• This technique holds promise for enhancing the precision and efficacy of image-guided radiation therapy.