Accelerated parallel magnetic resonance imaging with compressed sensing using structured sparsity.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Signal Processing

Background:

• Compressed sensing and parallel imaging are advanced MRI techniques.
• Model-based reconstruction methods have been used but do not fully exploit sparsity structures.
• Structured sparsity offers potential for enhanced image reconstruction.

Purpose of the Study:

• To develop and evaluate a method combining compressed sensing with parallel imaging that utilizes the structure of the sparsifying transformation.
• To improve Magnetic Resonance Imaging (MRI) reconstruction quality by incorporating structured sparsity.
• To reduce image reconstruction errors in MRI.

Main Methods:

• A novel method integrating compressed sensing with parallel imaging was developed.
• The approach takes advantage of the structure of the sparsifying transformation.
• An optimization problem was modified to incorporate blurry coil images from a fully sampled center region, estimating missing details.

Main Results:

• The combined method demonstrated lower relative error compared to sparse SENSE and L1 ESPIRiT.
• Reconstructions were performed using data from brain, ankle, and shoulder anatomies.
• The technique effectively utilized structured sparsity for improved image reconstruction.

Conclusions:

• Leveraging structured sparsity significantly enhances image quality for a given data acquisition.
• The method requires a fully sampled region centered on the zero frequency of adequate size.
• This approach offers a valuable improvement for MRI reconstruction, particularly in scenarios with limited data.