Motion prediction in MRI-guided radiotherapy based on interleaved orthogonal cine-MRI

Area of Science:

• Medical Physics
• Radiotherapy Technology
• Image-Guided Interventions

Background:

• In-room cine-magnetic resonance imaging (MRI) enables non-invasive radiotherapy target localization.
• Tumor motion prediction is crucial to manage imaging frequency and system latencies between localization and dose delivery.

Purpose of the Study:

• To propose a framework for tumor motion prediction specifically for cine-MRI guidance.
• To quantify geometric uncertainties in tumor tracking and beam gating introduced by motion prediction.

Main Methods:

• Tumor position identified via scale-invariant features in cine-MRI slices.
• High-frequency (25 Hz) position estimation using independent predictors for each coordinate.
• Comparison of linear extrapolation, auto-regressive, and support vector machine algorithms against no-prediction and surrogate-based methods.

Main Results:

• Tracking error root-mean-square (RMS) decreased to 0.2–1.2 mm with acquisition periods of 250–750 ms and system latencies of 50–300 ms.
• Prediction errors for tracking and gating were generally lower than surrogate-based methods, excluding the linear extrapolator.
• Geometric uncertainties were quantified as a function of image acquisition period and system latency.

Conclusions:

• Cine-MRI guidance combined with advanced prediction algorithms can significantly reduce geometric uncertainties in motion-compensated radiotherapy.
• The proposed framework demonstrates the potential for improved accuracy in tumor tracking and beam gating.
• Motion prediction is a key component for enhancing the efficacy of image-guided radiotherapy treatments.