Adaptive continuation based smooth l0-norm approximation for compressed sensing MR image reconstruction.

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Science

Background:

• Compressed Sensing (CS) Magnetic Resonance Image (MRI) reconstruction often relies on smooth L0-norm (SL0) approximation algorithms.
• Traditional methods use predefined parameter sequences, which may not yield optimal sparsity or reconstruction performance due to data-dependent parameter variations.

Purpose of the Study:

• To develop an adaptive compressed sensing MRI reconstruction method using SL0 approximation.
• To address the limitations of fixed parameter sequences by introducing a data-dependent adaptive approach.

Main Methods:

• Proposes an adaptive SL0 approximation algorithm for CS-MRI reconstruction.
• Employs an alternating strategy to solve sparse regularization and parameter estimation subproblems.
• Utilizes a root-finding technique for adaptive parameter estimation.

Main Results:

• The adaptive SL0 algorithm demonstrates significant improvements in speed and accuracy compared to existing methods.
• Achieved at least twofold faster CPU time than automated parameter estimation methods.
• Showcased an average 15% improvement in reconstruction performance (Normalized Mean Squared Error).

Conclusions:

• Presents a novel adaptive continuation-based SL0 algorithm for CS-MRI.
• This data-dependent method eliminates the need for searching optimal constant scale factor values.
• Offers a more efficient and accurate approach for CS-based MRI reconstruction.