Gaussianization of Diffusion MRI Data Using Spatially Adaptive Filtering.

Area of Science:

• Medical Imaging
• Biophysics
• Signal Processing

Background:

• Diffusion MRI data is susceptible to noise floor, impacting signal quality and diffusion parameter accuracy.
• Current phase correction methods often assume stationary noise, which is not always true for real-world diffusion MRI data.
• Accurate phase estimation is crucial for effective noise reduction in diffusion MRI.

Purpose of the Study:

• To develop an adaptive filtering approach for smoothing diffusion MRI images with spatially-varying noise.
• To improve the accuracy of phase correction in diffusion MRI by addressing the limitations of stationary noise assumptions.
• To enhance the signal Gaussianization process for more reliable diffusion parameter estimation.

Main Methods:

• Introduced the multi-kernel filter (MKF), an adaptive filtering technique utilizing a bilateral filter with spatially-varying kernels.
• Applied MKF for image smoothing to estimate background phase in complex diffusion-weighted images.
• Compared MKF performance against state-of-the-art filters in signal Gaussianization experiments.

Main Results:

• MKF demonstrated significant improvements in spatial adaptivity compared to existing methods.
• The proposed adaptive filtering approach effectively handles spatially-varying noise in diffusion MRI data.
• MKF outperformed other filters in achieving signal Gaussianization, reducing noise floor confounds.

Conclusions:

• The multi-kernel filter (MKF) offers a robust solution for noise reduction in diffusion MRI by adapting to spatially-varying noise.
• MKF enhances the accuracy of phase correction, leading to more reliable diffusion parameter estimation.
• This adaptive filtering approach holds promise for improving the quality and interpretability of diffusion MRI studies.