Particle Filter-Based Target Tracking Algorithm for Magnetic Resonance-Guided Respiratory Compensation: Robustness
Alexandra E Bourque1, Stéphane Bedwani1, Jean-François Carrier1
1Département de physique, Université de Montréal, Montréal, Québec, Canada; Département de radio-oncologie, Centre hospitalier de l'Université de Montréal, Montréal, Québec, Canada.
International Journal of Radiation Oncology, Biology, Physics
|November 22, 2017
Summary
A new particle filter algorithm accurately tracks tumors during MR-guided radiation therapy, showing high robustness against image quality variations for improved treatment accuracy.
Area of Science:
- Medical Physics
- Image-guided Therapy
- Computational Imaging
Background:
- Accurate tumor tracking is crucial for effective MR-guided radiation therapy.
- Existing tracking algorithms may struggle with variations in image quality and dynamic patient anatomy.
Purpose of the Study:
- To evaluate the robustness and accuracy of a modified particle filter algorithm for real-time tumor tracking.
- To assess the algorithm's performance across diverse magnetic resonance imaging (MRI) conditions.
Main Methods:
- Implemented an improved particle filter algorithm using normalized cross-correlation for likelihood calculation.
- Tested the algorithm on 24 dynamic MRI datasets from healthy volunteers and patients, varying resolution, contrast, and signal-to-noise ratio.
- Compared algorithm-derived tracking results against expert delineations to compute tracking errors.
Main Results:
- The modified algorithm successfully tracked both abdominal and thoracic tumors, outperforming a previous implementation that failed over 50% of the time.
- Achieved a mean tracking error of 1.1 ± 0.4 mm across all acquisitions, demonstrating significant robustness to image quality variations.
- Identified consistent effects of input/control parameters, suggesting potential for class-based optimization.
Conclusions:
- The enhanced particle filter tracking algorithm demonstrates high accuracy and robustness in the presence of varying image quality.
- This algorithm is a promising tool for automated tumor tracking in MR-guided radiation therapy, particularly on MR linear accelerators.
