Non-Stationary Rician Noise Estimation in Parallel MRI Using a Single Image: A Variance-Stabilizing Approach

Area of Science:

• Medical Imaging
• Signal Processing
• Magnetic Resonance Imaging

Background:

• Parallel magnetic resonance imaging (pMRI) accelerates image acquisition but introduces non-stationary noise.
• Existing noise estimation methods for pMRI often require unavailable data or make restrictive assumptions.
• Some current techniques result in undesirable granular patterns in reconstructed images.

Purpose of the Study:

• To develop an automatic noise estimation technique for non-stationary Rician noise in pMRI.
• To overcome limitations of existing methods, including data requirements and SNR dependency.
• To provide a robust solution applicable across various signal-to-noise conditions.

Main Methods:

• A novel variance-stabilizing transformation was derived to address non-stationary Rician noise.
• The proposed method is designed to function effectively across a wide range of signal-to-noise ratios (SNRs).
• The technique was validated against state-of-the-art methods using both synthetic and real pMRI data.

Main Results:

• The new noise estimation technique demonstrates robustness in various scenarios.
• The method shows superior performance compared to existing approaches for all tested SNRs.
• The technique effectively handles non-stationary Rician noise without requiring extensive prior information.

Conclusions:

• The proposed automatic noise estimation method offers a significant advancement for pMRI.
• The technique's effectiveness across all SNRs enhances its applicability in clinical and research settings.
• This method overcomes key limitations of prior noise estimation techniques in accelerated MRI.