Arterial spin labeling perfusion MRI signal denoising using robust principal component analysis

Area of Science:

• Medical Imaging
• Neuroscience
• Biophysics

Background:

• Arterial spin labeling (ASL) perfusion MRI non-invasively quantifies regional cerebral blood flow (CBF).
• ASL MRI is valuable for characterizing brain state changes in disease and functional alterations.
• Low signal-to-noise ratio (SNR) currently limits dynamic brain activity studies using ASL.

Purpose of the Study:

• To validate a novel temporal denoising strategy for ASL MRI.
• To improve the utility of ASL for dynamic brain activity studies.
• To enhance functional connectivity (FC) analysis and task activation detection using ASL data.

Main Methods:

• Employed robust principal component analysis (rPCA) to decompose ASL CBF image series into low-rank and sparse components.
• Developed a method to determine the rPCA parameter balancing low-rankness and sparsity, tailored for ASL MRI data structures.
• Evaluated the method using ASL CBF-based functional connectivity (FC) analysis and a sensorimotor functional ASL MRI study.
• Compared the proposed method against component based noise correction (CompCor).

Main Results:

• The proposed temporal denoising method significantly increased the temporal signal-to-noise-ratio (TSNR) of ASL CBF images.
• Demonstrated marked improvements in the sensitivity of ASL CBF images for FC analysis.
• Showcased enhanced sensitivity for task activation detection in functional ASL studies.

Conclusions:

• Introduced a novel temporal denoising technique for ASL CBF images.
• Validated the method's effectiveness in improving CBF time series-based FC analysis.
• Confirmed the method's benefits for enhancing task activation detection in ASL studies.