Adaptive noise filtering for accurate and precise diffusion estimation in fiber crossings

Area of Science:

• Medical Imaging
• Neuroimaging
• Diffusion MRI

Background:

• Diffusion Weighted MR Images (DMI) present challenges in measuring crossing fiber diffusion due to numerous parameters and low signal-to-noise ratio (SNR).
• Effective noise filtering is crucial for maintaining data distribution while suppressing noise in DMI.

Purpose of the Study:

• To develop an adaptive noise filtering technique for Diffusion Weighted MR Images.
• To enhance the precision and accuracy of diffusion property estimation, particularly for crossing fibers.

Main Methods:

• An adaptive version of the Linear Minimum Mean Square Error (LMMSE) estimator was proposed.
• The filter utilizes a space-variant noise level estimate and a diffusion similarity-based weighting kernel.
• The Rician data distribution is preserved post-filtering.

Main Results:

• The adaptive LMMSE filter demonstrated improved precision in diffusivity value estimation.
• The method maintained data accuracy compared to existing techniques.
• Experimental results on brain data showed superior performance over the standard LMMSE estimator.

Conclusions:

• The proposed adaptive LMMSE filter effectively reduces noise in Diffusion Weighted MR Images.
• This technique enhances the precise estimation of diffusion properties in complex white matter structures.
• The method offers a significant improvement for neuroimaging analyses involving crossing fibers.