Noise distribution in SENSE- and GRAPPA-reconstructed images: a computer simulation study

Area of Science:

• Magnetic Resonance Imaging (MRI)
• Image Reconstruction
• Medical Physics

Background:

• Parallel imaging techniques like SENSE and GRAPPA accelerate MRI acquisition.
• Understanding noise distribution is crucial for image quality assessment in accelerated MRI.
• Previous studies have focused on acceleration factors, but noise distribution differences require detailed investigation.

Purpose of the Study:

• To descriptively analyze and compare noise distributions in images reconstructed using SENSE and GRAPPA.
• To evaluate the impact of regularization on SENSE noise patterns.
• To investigate noise characteristics across different reduction factors in parallel imaging.

Main Methods:

• Computer simulations were performed using synthetic and in vitro objects.
• Coil sensitivities and k-space data were simulated with added noise.
• Image reconstruction was conducted using SENSE (with and without regularization) and GRAPPA at various reduction factors.

Main Results:

• Significant differences in noise distributions were observed between SENSE- and GRAPPA-reconstructed images.
• GRAPPA reconstructions exhibited a more spatially uniform noise distribution compared to SENSE.
• SENSE reconstructions showed noise variations influenced by the geometry factor, with regularization reducing but not eliminating this pattern.

Conclusions:

• GRAPPA demonstrates a more homogeneous noise profile, potentially offering advantages in certain clinical applications.
• SENSE noise characteristics are dependent on coil geometry, a factor mitigated but not entirely removed by regularization.
• The findings provide valuable insights into the trade-offs between SENSE and GRAPPA regarding noise performance in parallel MRI.