Comparison and analysis of nonlinear algorithms for compressed sensing in MRI

Area of Science:

• Medical Imaging
• Signal Processing
• Computational Science

Background:

• Compressed sensing (CS) theory accelerates Magnetic Resonance Imaging (MRI).
• Various algorithms exist to solve CS reconstruction problems, but optimal selection criteria for MRI are lacking.
• A systematic comparison of CS algorithms is crucial for practical MRI implementation.

Purpose of the Study:

• To systematically compare the performance of three common compressed sensing algorithms in MRI.
• To evaluate algorithms based on image quality and reconstruction speed across different imaging scenarios.
• To provide insights for selecting optimal CS algorithms in MRI applications.

Main Methods:

• Comparison of Gradient Projection For Sparse Reconstruction (GPSR), Interior-point (l(1)_ls), and Stagewise Orthogonal Matching Pursuit (StOMP) algorithms.
• Investigation across three distinct imaging scenarios: brain, angiogram, and phantom imaging.
• Performance characterization based on image quality metrics and reconstruction speed.

Main Results:

• Algorithm performance is highly case-sensitive in MRI applications.
• The Stagewise Orthogonal Matching Pursuit (StOMP) algorithm demonstrated superior image quality.
• The Gradient Projection For Sparse Reconstruction (GPSR) algorithm exhibited the highest reconstruction efficiency.

Conclusions:

• The choice of CS algorithm significantly impacts MRI performance (image quality vs. speed).
• StOMP and GPSR present distinct advantages for different MRI application needs.
• Further experimental validation is planned to explore algorithm characteristics in practice, supporting CS adoption in MRI.