K-space reconstruction with anisotropic kernel support (KARAOKE) for ultrafast partially parallel imaging

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Medical Imaging Physics
• Image Reconstruction Algorithms

Background:

• Partially parallel imaging (PPI) accelerates MRI using surface coil arrays and k-space undersampling.
• The theoretical acceleration factor (R) in PPI is limited by the number of coils (N(C)).
• Symmetric kernels often prevent achieving the theoretical R, impacting image quality.

Purpose of the Study:

• To develop a kernel design method for accelerating PPI beyond the theoretical R = N(C) constraint.
• To introduce a novel approach for faster and higher-quality MR image reconstruction.

Main Methods:

• Exploited k-space data anisotropy, correlated with object and coil sensitivity profiles.
• Proposed asymmetric kernels derived from fitting bivariate Gaussian functions to local signal magnitude.
• Utilized a perceptual difference model (Case-PDM) for quantitative image quality evaluation.

Main Results:

• K-space anisotropy increases with magnetic field strength.
• The KARAOKE algorithm demonstrated superior edge preservation compared to GRAPPA, especially at high R.
• KARAOKE outperformed GRAPPA at high R (near or exceeding N(C)), while performing comparably at low R.

Conclusions:

• KARAOKE reconstruction offers higher quality k-space reconstruction.
• Recommended for PPI data acquired at high acceleration factors (R) and/or high magnetic field strengths.
• Enhances image quality for applications like cardiac imaging and motion correction.