Undersampled MRI reconstruction based on spectral graph wavelet transform.

Area of Science:

• Medical Imaging
• Signal Processing
• Computer Vision

Background:

• Compressed sensing magnetic resonance imaging (CS-MRI) offers accelerated imaging but relies on effective sparse image representation.
• The quality of image reconstruction in CS-MRI is highly dependent on the chosen sparsifying transform.

Purpose of the Study:

• To introduce and evaluate a spectral graph wavelet transform (SGWT) for sparse representation in CS-MRI.
• To demonstrate the effectiveness of SGWT in improving iterative image reconstruction quality.

Main Methods:

• Developed SGWT by extending traditional wavelets to signals on weighted graphs, utilizing connectivity information.
• Implemented a Chebyshev polynomial approximation for fast SGWT computation.
• Utilized an l1-norm regularized CS-MRI reconstruction model solved via projected iterative soft-thresholding.

Main Results:

• The proposed SGWT method effectively sparsifies magnetic resonance images for iterative reconstruction.
• Numerical experiments showed SGWT outperforms state-of-the-art sparsifying transforms.
• SGWT demonstrated superior performance in artifact suppression and achieving lower reconstruction errors.

Conclusions:

• SGWT provides a robust and effective method for sparse representation in CS-MRI.
• The developed SGWT is applicable to any domain representable by a weighted graph.
• This approach significantly enhances CS-MRI reconstruction quality and efficiency.