Spatially variable Rician noise in magnetic resonance imaging

Area of Science:

• Medical Imaging
• Biophysics
• Signal Processing

Background:

• Magnetic resonance images (MRI) are susceptible to noise from hardware and physiological sources, impacting image resolution and reproducibility.
• Noise reduction is crucial for MRI, especially in low signal-to-noise ratio applications like diffusion tensor imaging (DTI).
• Current noise correction methods often assume uniform noise distribution, limiting their effectiveness.

Purpose of the Study:

• To develop and validate an advanced noise reduction technique for MRI.
• To address the limitations of standard methods by considering spatially inhomogeneous noise distribution.
• To improve image quality and reliability in DTI and other MRI applications.

Main Methods:

• Developed a novel noise correction approach evaluating individual correction factors for each voxel.
• Incorporated the Rician distribution model to account for noise characteristics at varying signal-to-noise ratios.
• Validated the method using numerical simulations and in vivo human brain DTI data.

Main Results:

• The proposed method effectively reduces noise by considering spatially varying noise levels.
• Demonstrated improved image resolution and reproducibility compared to standard methods.
• The technique proved effective for low signal-to-noise ratio images, such as those in DTI.

Conclusions:

• The developed voxel-wise noise correction method offers a significant improvement over standard techniques.
• This approach enhances the performance of clinical and research MRI investigations, particularly DTI.
• The validated tool provides effective noise reduction for improved diagnostic accuracy and data reliability.