Iterative Schemes to Solve Low-Dimensional Calibration Equations in Parallel MR Image Reconstruction with GRAPPA

Area of Science:

• Magnetic Resonance Imaging
• Image Reconstruction
• Computational Imaging

Background:

• Generalized Autocalibrating Partially Parallel Acquisition (GRAPPA) is a standard parallel MRI reconstruction technique.
• High channel counts in MRI increase computational and storage demands for GRAPPA.
• Random Projection GRAPPA (RP-GRAPPA) reduces computational load but can decrease accuracy.

Purpose of the Study:

• To implement GRAPPA reconstruction using iterative solvers for enhanced accuracy with random projection.
• To evaluate the performance of iterative solvers against RP-GRAPPA, particularly with significant dimension reduction.
• To assess computational time savings and accuracy preservation.

Main Methods:

• Implementation of GRAPPA reconstruction utilizing iterative solvers (e.g., CGLS) on randomly projected calibration data.
• Comparison of proposed iterative methods with standard RP-GRAPPA.
• Quantitative and visual assessment of reconstruction accuracy and computational efficiency.

Main Results:

• Proposed iterative methods maintain reconstruction accuracy comparable to RP-GRAPPA, even with large dimension reductions.
• Conjugate Gradient for Least Squares (CGLS) demonstrated significant computational time savings over RP-GRAPPA.
• The CGLS method complements channel compression techniques, reducing memory usage and increasing speed for high-channel-count systems.

Conclusions:

• Iterative solvers offer a viable solution to improve RP-GRAPPA accuracy.
• CGLS provides an efficient and accurate method for GRAPPA reconstruction, especially for high-channel-count MRI.
• The proposed approach enhances the practicality of GRAPPA for demanding imaging scenarios.