Trimmed autocalibrating k-space estimation based on structured matrix completion.

Area of Science:

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

Background:

• Parallel imaging reconstructs undersampled MRI data using multiple coils.
• Corrupted data, such as from motion artifacts, poses challenges for standard reconstruction.
• Existing methods struggle with calibration data corrupted by artifacts.

Purpose of the Study:

• To adapt the SAKE (Structured low rank approximation based matrix completion) parallel imaging method for reconstructing MRI data with corrupted calibration information.
• To develop a robust method capable of detecting and removing erroneous data points within the calibration region of k-space.

Main Methods:

• The SAKE method, originally based on matrix completion and structured low rank approximation, was modified.
• The modification involved incorporating a data trimming strategy to identify and exclude corrupted data points.
• Reconstructions were performed using the modified SAKE, a standard linear equation solver with data rejection, and simulated/actual corrupted datasets.

Main Results:

• The modified SAKE method with data trimming demonstrated superior performance in reducing artifacts.
• Compared to a standard reconstruction method, the proposed approach yielded significantly cleaner images.
• Both simulated and real-world corrupted datasets showed considerable improvement with the SAKE-based method.

Conclusions:

• SAKE parallel imaging, augmented with data trimming, offers an improved solution for reconstructing MRI images from severely corrupted datasets.
• This enhanced method effectively handles artifacts that compromise standard calibration and reconstruction techniques.
• The findings highlight the potential of SAKE for robust image reconstruction in challenging MRI scenarios.