Accelerating volumetric cine MRI (VC-MRI) using undersampling for real-time 3D target localization/tracking in...

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Computational Anatomy

Background:

• Accelerating volumetric cine MRI (VC-MRI) is crucial for real-time applications in radiotherapy.
• Current methods often require significant acquisition time, limiting their clinical utility.
• Developing faster VC-MRI techniques can enhance treatment accuracy and patient outcomes.

Purpose of the Study:

• To accelerate VC-MRI acquisition by utilizing undersampled 2D-cine MRI data.
• To provide real-time 3D guidance for gating and target tracking during radiotherapy.
• To evaluate the feasibility and accuracy of the proposed acceleration method.

Main Methods:

• Employed principal component analysis (PCA) to model tissue motion from prior 4D-MRI data.
• Reconstructed VC-MRI by solving for motion coefficients that best match undersampled on-board 2D-cine MRI.
• Investigated both Cartesian and radial k-space sampling strategies with varying sampling percentages (SP).
• Assessed the impact of tumor size and noise on VC-MRI estimation accuracy.

Main Results:

• XCAT simulations showed minimal volume percent difference (VPD) and center of mass shift (COMS) with as low as 5.2% radial or 10% Cartesian sampling.
• Accuracy improved with higher sampling rates for smaller tumors.
• VC-MRI estimation demonstrated robustness against noise up to a signal-to-noise ratio (SNR) of 20.
• Patient data yielded tumor tracking errors below 1 mm in all directions for both sampling methods.

Conclusions:

• Feasible to estimate VC-MRI from single, undersampled 2D cine MRI acquisitions.
• Significant improvement in temporal resolution (5-10x) achievable with 10-20% k-space sampling.
• This accelerated VC-MRI holds promise for real-time guidance in radiotherapy.