Exploiting rank deficiency and transform domain sparsity for MR image reconstruction

Area of Science:

• Medical Imaging
• Signal Processing
• Applied Mathematics

Background:

• Compressed sensing (CS) methods reconstruct magnetic resonance (MR) images from partial k-space data by exploiting image sparsity.
• Recent work demonstrates MR image reconstruction is possible by leveraging image rank deficiency.
• Existing methods individually exploit either sparsity or rank deficiency for MR image reconstruction.

Purpose of the Study:

• To investigate the combined benefits of transform domain sparsity and rank deficiency for MR image reconstruction.
• To develop a novel optimization framework that integrates both sparsity and rank deficiency principles.
• To propose and validate a new algorithm for solving the combined minimization problem.

Main Methods:

• Formulated a novel optimization problem combining L1-norm minimization (for transform domain sparsity) and nuclear norm minimization (for rank deficiency).
• Developed and derived a first-order algorithm to solve this previously unencountered optimization problem.
• Evaluated reconstruction performance using visual quality and signal-to-noise ratio metrics.

Main Results:

• The proposed method, combining sparsity and rank deficiency, achieved superior MR image reconstruction.
• Significant improvements in visual quality were observed compared to methods using only sparsity or only rank deficiency.
• Enhanced signal-to-noise ratio was demonstrated, indicating improved data fidelity.

Conclusions:

• Combining transform domain sparsity with rank deficiency offers a powerful approach for MR image reconstruction.
• The proposed L1-norm and nuclear norm minimization framework effectively leverages both image properties.
• This novel method provides substantial improvements over existing CS-based and rank deficiency-based techniques for MR imaging.