Related Experiment Video
Updated: Jul 28, 2025
Dynamic Lung Tumor Tracking for Stereotactic Ablative Body Radiation Therapy
Published on: June 7, 2015
Gaussian Processes for real-time 3D motion and uncertainty estimation during MR-guided radiotherapy
Niek R F Huttinga1, Tom Bruijnen1, Cornelis A T van den Berg1
1Department of Radiotherapy, Division of Imaging & Oncology, University Medical Center Utrecht, The Netherlands; Computational Imaging Group for MR diagnostics & therapy, Center for Image Sciences, University Medical Center Utrecht, The Netherlands.
This study introduces a real-time framework using Gaussian Processes to accurately track respiratory motion during MR-guided radiotherapy (MRgRT). The system provides motion uncertainty maps, enhancing patient safety and treatment efficacy.
Area of Science:
- Medical Physics
- Radiotherapy Technology
- Medical Imaging
Background:
- Respiratory motion introduces significant uncertainty in tumor targeting during radiotherapy, often necessitating larger radiation margins and reduced doses, thereby compromising treatment efficacy.
- Hybrid MR-linac systems offer real-time adaptive MR-guided radiotherapy (MRgRT) potential for managing respiratory motion.
- Accurate and rapid estimation of 3D motion fields from MR data, along with a confidence measure, is crucial for safe and effective real-time MRgRT.
Purpose of the Study:
- To develop and validate a real-time framework for inferring 3D respiratory motion fields and their uncertainties from limited MR data.
- To achieve a total latency of under 200 ms for motion estimation and adaptation in MR-guided radiotherapy.
- To establish a quality assurance measure for motion estimation confidence to ensure patient safety.
Main Methods:
- A novel framework utilizing Gaussian Processes was developed to infer 3D motion fields and uncertainty maps from minimal MR data (three readouts).
- The framework's inference speed was optimized, achieving frame rates up to 69 Hz, including data acquisition and reconstruction.
- A rejection criterion based on motion-field uncertainty maps was designed for quality assurance.
Main Results:
- The framework demonstrated high-speed inference (up to 69 Hz) suitable for real-time MR-guided radiotherapy applications.
- In silico validation showed end-point errors with a 75th percentile below 1 mm.
- The rejection criterion effectively detected erroneous motion estimates, confirming the framework's potential for quality assurance in vivo.
Conclusions:
- The proposed Gaussian Process-based framework enables real-time 3D motion field and uncertainty estimation from limited MR data.
- The framework meets the stringent latency requirements for real-time adaptive MR-guided radiotherapy on MR-linac systems.
- The developed uncertainty maps and rejection criterion provide a valuable tool for quality assurance, enhancing patient safety in MR-guided radiotherapy.
