A cosparse analysis model with combined redundant systems for MRI reconstruction.

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Science

Background:

• Magnetic Resonance Imaging (MRI) is a crucial noninvasive diagnostic tool.
• MRI scans typically involve long acquisition times, limiting patient throughput and comfort.
• Accelerating MRI data acquisition without compromising image quality is a significant challenge.

Purpose of the Study:

• To develop an advanced method for reconstructing high-quality MR images from undersampled k-space data.
• To accelerate MRI data acquisition through efficient image reconstruction techniques.
• To improve the diagnostic utility of MRI by enhancing image fidelity from sparse sampling.

Main Methods:

• A cosparse analysis model is proposed, utilizing combined redundant systems for enhanced sparsity exploitation.
• Wavelet tight frames and Gabor frames are employed to characterize diverse image structures.
• An alternating iterative reconstruction scheme is implemented for efficient and effective image recovery.

Main Results:

• The proposed reconstruction method demonstrates superior performance compared to existing state-of-the-art techniques.
• Evaluations on brain MR images show significant improvements in Peak Signal-to-Noise Ratio (PSNR), exceeding 9 dB under 10% random sampling.
• The method achieves high-quality reconstructions across various sampling ratios (10%–60%) and sampling patterns.

Conclusions:

• The cosparse analysis model effectively characterizes image sparsities using combined wavelet and Gabor frames.
• Partial norm regularization aids in obtaining optimal solutions in lower dimensions.
• The proposed method significantly improves MRI reconstruction quality, particularly for highly undersampled datasets.