Balanced sparse model for tight frames in compressed sensing magnetic resonance imaging.

Area of Science:

• Medical Imaging
• Signal Processing
• Optimization Algorithms

Background:

• Compressed sensing (CS) is a key technology for accelerating magnetic resonance imaging (MRI).
• Image reconstruction in CS-MRI typically relies on sparsity enforcement using synthesis or analysis models.
• The balanced model offers a novel approach by integrating aspects of both synthesis and analysis models.

Purpose of the Study:

• To evaluate the performance of the balanced model in tight frame-based CS-MRI.
• To introduce an efficient numerical algorithm for solving the balanced model's optimization problem.
• To compare the balanced model's performance against existing synthesis and analysis models.

Main Methods:

• Implementation and analysis of the balanced sparse model for compressed sensing MRI.
• Development of a novel constrained split augmented Lagrangian shrinkage algorithm for the balanced model (C-SALSA-B).
• Comparative performance evaluation against Accelerated Proximal Gradient (APG) and Alternating Directional Method of Multipliers (ADMM) algorithms.

Main Results:

• The balanced model, by tuning its parameter, can achieve solutions comparable to synthesis and analysis models.
• The balanced model consistently outperforms the synthesis model.
• The proposed C-SALSA-B algorithm demonstrates faster convergence than APG and ADMM-B for the balanced model.

Conclusions:

• The balanced model presents a robust framework for compressed sensing MRI reconstruction.
• The C-SALSA-B algorithm provides an efficient and faster solution for balanced model optimization.
• The balanced model offers comparable or superior performance to existing sparse reconstruction models in CS-MRI.