Patch tensor decomposition and non-local means filter-based hybrid ASL image denoising

Area of Science:

• Neuroimaging
• Medical Physics
• Biomedical Engineering

Background:

• Arterial spin labeling magnetic resonance imaging (ASL MRI) is a noninvasive technique for measuring cerebral blood flow (CBF), crucial for studying neurodegenerative diseases.
• ASL MRI images have a low signal-to-noise ratio (SNR), necessitating effective image denoising for accurate analysis.
• Image denoising is a critical preprocessing step in ASL MRI to improve the quality of cerebral blood flow (CBF) perfusion images.

Purpose of the Study:

• To introduce a novel ASL image denoising method.
• To evaluate the proposed method against existing ASL denoising techniques.
• To demonstrate the enhanced performance of the denoising method in functional connectivity and task activation analyses.

Main Methods:

• The proposed method utilizes patch-based low-rank and sparse tensor decomposition combined with a non-local means filter.
• Comparison was made with two established methods: component-based noise correction (CompCor) and low-rank and sparse matrix decomposition (LS-ASLd).

Main Results:

• The proposed method demonstrated superior performance in improving SNR, tSNR, and reducing ASL CBF variance compared to existing methods.
• Denoised images from resting-state ASL datasets showed enhanced functional connectivity (FC) analysis.
• Denoised images from task-related ASL datasets revealed improved brain activation detection.

Conclusions:

• The developed hybrid ASL CBF image denoising method is highly effective.
• The method is particularly suitable for enhancing functional connectivity (FC) analysis.
• The denoising technique shows significant utility in sensorimotor task analysis using ASL MRI data.