Incorporating reference in parallel imaging and compressed sensing.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Image Reconstruction
• Compressed Sensing

Background:

• Parallel imaging techniques accelerate MRI acquisition by using multiple receiver coils.
• Compressed sensing (CS) enables reconstruction from undersampled data by exploiting signal sparsity.
• Integrating prior information can further improve reconstruction quality and reduce artifacts.

Purpose of the Study:

• To develop a novel compressed sensing parallel imaging (CSPI) technique, READ-PICS (Reference-Enhanced Adaptive Dictionary-based Parallel Imaging Compressed Sensing).
• To leverage prior information from a reference scan for enhanced MR image reconstruction from highly undersampled multichannel data.
• To simultaneously increase image sparsity and mitigate noise amplification during reconstruction.

Main Methods:

• READ-PICS utilizes a high-spatial-resolution reference prior via the generalized series model.
• Central k-space residual calibration is employed to address parallel imaging system ill-conditioning.
• Joint sparsity is enforced using a mixed L1-L2 norm on prior and residual component coefficients.

Main Results:

• READ-PICS demonstrated superior performance in detail preservation and noise suppression compared to L1-Iterative self-consistent parallel imaging reconstruction and correlation imaging.
• Evaluations on parametric imaging and multiscan experiments validated the technique's effectiveness.
• The method significantly improved image sparsity and addressed ill-conditioning in parallel imaging.

Conclusions:

• The proposed READ-PICS technique effectively enhances signal sparsity and improves parallel imaging system conditioning through reference adaptive regularization.
• This method offers a robust approach for MR image reconstruction from undersampled multichannel data.
• READ-PICS is adaptable to other imaging applications requiring sequential acquisition with available reference priors.