Motion-aligned 4D-MRI reconstruction using higher degree total variation and locally low-rank regularization

Area of Science:

• Medical Imaging
• Radiotherapy Physics
• Image Reconstruction

Background:

• Four-dimensional magnetic resonance imaging (4D-MRI) offers non-ionizing 3D structural and temporal data crucial for radiotherapy planning.
• Long acquisition times and motion artifacts currently limit the clinical utility of 4D-MRI.

Purpose of the Study:

• To develop an accelerated 4D-MRI reconstruction method addressing limitations of long acquisition times and motion artifacts.
• To improve image quality and reduce artifacts in 4D-MRI for enhanced radiotherapy treatment planning.

Main Methods:

• Proposed a novel motion-aligned reconstruction method (maHDTV-LLR) using higher degree total variation and locally low-rank regularization.
• Implemented a two-stage framework alternating motion alignment and regularized optimization.
• Integrated 3D-HDTV and locally low-rank penalties to leverage spatial and temporal correlations.
• Employed a fast alternating minimization algorithm with variable splitting for efficient optimization.

Main Results:

• Successfully reconstructed 4D MR images from highly undersampled Fourier coefficients.
• Demonstrated significant improvements in image quality and reduction of artifacts in cardiac and abdominal 4D-MRI.
• Validated the method's effectiveness at high undersampling factors.

Conclusions:

• The proposed maHDTV-LLR method enables accelerated 4D-MRI acquisition.
• This technique enhances image quality and reduces artifacts, paving the way for broader clinical application in radiotherapy.
• The method effectively exploits spatial and temporal correlations in 4D-MRI data.