Well-designed k-space coverage is important for good MRI denoising.

Area of Science:

• Medical Imaging
• Computational MRI
• Signal Processing

Background:

• Modern MRI denoising often assumes fixed k-space coverage.
• Prior work explored modifying k-space coverage (e.g., reducing resolution) to boost SNR.
• This study examines if k-space modifications benefit current denoising techniques.

Purpose of the Study:

• To investigate the impact of k-space coverage modifications on advanced MRI denoising.
• To determine if optimizing k-space patterns enhances denoising performance.
• To compare optimized vs. naive coverage strategies for denoising.

Main Methods:

• Simulated noisy MRI data were used for optimization.
• K-space coverage and averaging patterns were optimized for U-Net and parallel imaging with total variation regularization.
• Performance was assessed using normalized root-mean-squared error (NRMSE) and structural similarity (SSIM).

Main Results:

• Reducing spatial resolution in MRI acquisition substantially improves denoising performance (quantitative and qualitative).
• Optimized k-space coverage with linear filtering/apodization achieved results comparable to advanced methods using higher-resolution naive coverage.
• Signal-to-noise ratio (SNR) gains from resolution reduction are key to enhanced denoising.

Conclusions:

• Traditional MRI acquisition principles (trading resolution for SNR) remain relevant for computational denoising.
• Optimization of k-space coverage and denoising methods can be complex.
• NRMSE and SSIM metrics show limited sensitivity to spatial resolution changes, complicating optimization.