Calibration-Less Multi-Coil Compressed Sensing Magnetic Resonance Image Reconstruction Based on OSCAR Regularization

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging (MRI)
• Signal Processing

Background:

• Compressed sensing (CS) combined with multi-coil acquisition accelerates MRI scans while maintaining signal-to-noise ratio (SNR).
• Self-calibrating techniques like ESPiRIT are standard for estimating coil sensitivity maps before reconstruction.
• Existing calibration-less methods often involve alternating sensitivity map and image reconstruction, which can be non-convex.

Purpose of the Study:

• Introduce a novel calibration-less multi-coil CS reconstruction method for MRI.
• Address the non-convexity issue in calibration-less reconstruction by reconstructing individual coil images.
• Improve image reconstruction quality and preserve phase information in accelerated MRI scans.

Main Methods:

• Reconstruct individual complex-valued MR images for each coil to avoid non-convexity.
• Utilize structured sparsity in the wavelet domain to compensate for ill-posedness.
• Employ OSCAR (octagonal shrinkage and clustering algorithm for regression) regularization to adapt to varying SNR across coils.
• Minimize a convex, non-smooth objective function using the proximal primal-dual Condat-Vù algorithm.

Main Results:

• The proposed method significantly outperforms state-of-the-art techniques (ℓ1-ESPIRiT, AC-LORAKS, CaLM) on magnitude images for T1 and FLAIR contrasts in retrospective studies.
• Validation on prospective 7 Tesla ex vivo human brain MRI data (8-20x acceleration) confirms retrospective findings.
• OSCAR-based regularization demonstrates superior preservation of phase information, both visually and quantitatively, compared to other methods.

Conclusions:

• The developed calibration-less multi-coil CS reconstruction method offers significant improvements in MRI.
• The approach enhances image quality and accurately preserves phase information, particularly beneficial for advanced MRI applications.
• This method represents a valuable advancement for accelerated, high-resolution MRI acquisition and reconstruction.