A maximum likelihood approach to parallel imaging with coil sensitivity noise.

Area of Science:

• Medical Imaging
• Magnetic Resonance Imaging
• Signal Processing

Background:

• Parallel imaging accelerates magnetic resonance (MR) acquisition using multiple coils.
• Coil sensitivity maps are crucial for reconstruction but often contain errors.
• Existing methods like SENSE assume noiseless sensitivity maps, which is unrealistic.

Purpose of the Study:

• To develop a robust parallel imaging reconstruction method that accounts for sensitivity map errors.
• To address the errors-in-variables problem arising from sensitivity noise in parallel imaging.
• To improve image quality and reconstruction performance under non-ideal conditions.

Main Methods:

• A maximum likelihood approach was used to model independent Gaussian sensitivity noise.
• A quasi-quadratic objective function was derived and efficiently minimized.
• The proposed method was evaluated against conventional SENSE.

Main Results:

• The new method demonstrated substantial gains over conventional SENSE.
• Significant improvements were observed in low signal-to-noise ratio (SNR) conditions.
• Performance enhancements were noted with high g-factors, large acceleration, and imperfect sensitivity maps (misalignment, ringing, noise).

Conclusions:

• The proposed maximum likelihood method effectively handles sensitivity map errors in parallel imaging.
• This technique offers superior performance compared to traditional SENSE, particularly in challenging MR imaging scenarios.
• The findings suggest a more robust and accurate approach for parallel MR image reconstruction.