Multi-channel framelet denoising of diffusion-weighted images.

Area of Science:

• Medical Imaging
• Biomedical Engineering
• Image Processing

Background:

• Diffusion MRI (DW-MRI) is crucial for microstructural analysis but suffers from low signal-to-noise ratio (SNR).
• Traditional denoising methods like total variation (TV) can introduce undesirable stair-casing artifacts.
• Existing methods often denoise individual diffusion-weighted (DW) images separately, limiting noise reduction efficacy.

Purpose of the Study:

• To develop an edge-preserving denoising technique for DW-MRI data.
• To overcome the limitations of TV-based denoising, specifically stair-casing effects.
• To improve the SNR of DW-MRI images for more accurate microstructural analysis.

Main Methods:

• A tight wavelet frame approach utilizing the unitary extension principle (UEP) to create discrete differential operators.
• Collaborative denoising of DW images acquired with adjacent gradient directions.
• An efficient ℓ0 denoising method involving thresholding and solving an inverse problem.

Main Results:

• The proposed method effectively preserves edges in DW-MRI data.
• Stair-casing artifacts commonly seen in TV denoising are significantly reduced.
• Qualitative and quantitative evaluations on synthetic and real data demonstrate the method's effectiveness.

Conclusions:

• The tight wavelet frame approach offers superior edge-preserving denoising for DW-MRI compared to traditional methods.
• Collaborative denoising and the efficient ℓ0 solver enhance image quality and analytical accuracy.
• This method holds promise for improving the reliability of microstructural information derived from DW-MRI.