Nonrigid 3D motion estimation at high temporal resolution from prospectively undersampled k-space data using low-rank...

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Anatomy

Background:

• The integration of Magnetic Resonance Imaging (MRI) with linear accelerators (LINAC) for radiotherapy necessitates advanced motion management techniques.
• Accurate characterization of internal body motion, including tumors and organs-at-risk, is crucial for effective radiotherapy planning and delivery.
• Retrospective analysis of motion from undersampled k-space data is essential for improving MR-guided radiotherapy.

Purpose of the Study:

• To introduce low-rank MR-MOTUS, a framework for retrospective reconstruction of time-resolved nonrigid 3D+t motion fields.
• To enable motion estimation from a single low-resolution reference image and prospectively undersampled k-space data.
• To facilitate the characterization of tumor and organs-at-risk motion during radiotherapy.

Main Methods:

• Utilizing a low-rank motion model to exploit spatiotemporal correlations in internal body motion.
• Inverting a signal model that directly relates motion fields to a reference image and k-space data.
• Applying the framework to 2D/3D abdominothoracic and 3D head scans acquired with golden-ratio radial readouts.

Main Results:

• Demonstrated retrospective estimation of 2D+t respiratory motion at 40.8 motion fields per second.
• Achieved 3D+t respiratory motion estimation at 7.6 motion fields per second and 3D+t head-neck motion at 9.3 motion fields per second.
• Validated reconstructed motion fields against time-resolved and static image reconstructions, showing good consistency.

Conclusions:

• The MR-MOTUS framework successfully estimates time-resolved nonrigid 3D motion fields.
• This capability allows for detailed characterization of breathing motion patterns during radiotherapy.
• The framework holds potential for real-time MR-guided radiotherapy applications.