A motion prediction confidence estimation framework for prediction-based radiotherapy gating.

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Image-Guided Interventions

Background:

• Motion prediction is crucial for mitigating latency in image-guided radiotherapy.
• Existing motion prediction methods are susceptible to errors and variations, impacting treatment efficacy.
• Adaptive gating in Magnetic Resonance Imaging (MRI)-guided radiotherapy presents a specific challenge for motion prediction accuracy.

Purpose of the Study:

• To develop and evaluate a novel motion prediction confidence estimation framework.
• To enhance the efficacy and robustness of prediction-based radiotherapy gating decisions.
• To generalize the method for various modalities and motion management setups.

Main Methods:

• A weighted combination of prediction model's goodness of fit, leave-one-out prediction variation, and target velocity was used.
• Weight parameters and action levels for beam-off decisions were optimized.
• The framework was validated in 8 healthy volunteer and 13 patient studies using a 0.35T MRI-guided radiotherapy system.

Main Results:

• Confidence estimator thresholding improved gating accuracy by up to 2.42% and increased positive predictive value (PPV) by up to 3.00%.
• Median centroid distance was reduced by up to 0.28 mm, with average accuracy reaching 96.5% and PPV 96.7%.
• The combination of goodness of fit and leave-one-out variation proved most effective, outperforming methods including target velocity.

Conclusions:

• Confidence estimation, using model fitness and validation, complements prediction methods for MRI-guided radiotherapy gating.
• The developed framework demonstrated improved gating quality in both volunteer and patient studies.
• This approach offers a more robust solution for prediction-based radiotherapy gating decisions.