Dynamic MRI reconstruction from highly undersampled (k, t)-space data using weighted Schatten p-norm regularizer of...

Area of Science:

• Medical Imaging
• Image Reconstruction
• Magnetic Resonance Imaging

Background:

• Dynamic MRI reconstruction often uses low-rank and sparsity assumptions.
• Existing methods using nuclear norm may deviate from true low-rank solutions.
• Current approaches lack flexibility in handling different rank components.

Purpose of the Study:

• To propose an efficient and flexible reconstruction model for dynamic MRI.
• To address limitations of convex relaxations in low-rank MRI reconstruction.
• To improve the quality of dynamic MRI images reconstructed from undersampled data.

Main Methods:

• Dynamic MRI data treated as a 3rd-order tensor.
• Low-rankness formulated using a weighted, non-convex Schatten p-norm.
• Reconstruction model combines weighted Schatten p-norm and spatiotemporal total variation.
• Algorithm based on Bregman iterations with alternating direction multiplier.

Main Results:

• The proposed method significantly enhances image quality in dynamic MRI.
• Experiments on public datasets validate the effectiveness of the new model.
• Weighted Schatten p-norm offers more flexibility than traditional nuclear norm.

Conclusions:

• The developed weighted Schatten p-norm regularizer improves dynamic MRI reconstruction.
• The proposed model provides a more accurate and flexible approach to low-rank tensor completion.
• This method offers superior performance for reconstructing dynamic MRI from undersampled k,t-space data.