Model-based iterative reconstruction for single-shot EPI at 7T.

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Image Reconstruction
• Signal Processing

Background:

• Echo planar imaging (EPI) is susceptible to artifacts like k-space nonuniformity, Nyquist ghosting, and geometric distortions.
• These artifacts arise from factors such as ramp sampling and B0 field inhomogeneity, complicating image reconstruction.
• Conventional methods often address these issues post-reconstruction or require additional calibration scans.

Purpose of the Study:

• To introduce a model-based reconstruction strategy for single-shot EPI.
• The strategy aims to intrinsically correct for k-space nonuniformity, Nyquist ghosting, and geometric distortions during reconstruction.
• The goal is to enhance EPI image quality without necessitating pre- or post-reconstruction processing steps.

Main Methods:

• Employs a nonuniform fast Fourier transform to handle non-Cartesian EPI data grids caused by ramp sampling and B0 inhomogeneity.
• Integrates a 2D Nyquist ghost phase correction directly into the reconstruction, eliminating the need for extra navigator acquisitions.
• Incorporates coil compression to decrease computational demands, with application to phantom and human brain MRI data.

Main Results:

• Demonstrates significant reduction in Nyquist ghosting and geometric distortions.
• The proposed 2D phase correction outperforms conventional 1D correction methods.
• Achieved improved temporal signal-to-noise ratio (tSNR) due to artifact reduction.
• Coil compression with 8 virtual coils reduced processing time by up to 75% with minimal tSNR loss (3.2%) for twofold undersampled data.

Conclusions:

• The developed reconstruction method enhances EPI image quality, addressing ghosting, geometry, and tSNR.
• It successfully corrects artifacts without requiring calibration data for Nyquist ghost correction.
• This approach offers a more efficient and robust method for EPI reconstruction.