GLOBAL PDF-BASED TEMPORAL NON-LOCAL MEANS FILTERING REVEALS INDIVIDUAL DIFFERENCES IN BRAIN CONNECTIVITY

Area of Science:

• Neuroimaging
• Biophysics
• Signal Processing

Background:

• Resting fMRI is crucial for brain connectivity analysis but faces challenges from low signal-to-noise ratio (SNR) and BOLD signal contrast.
• Traditional Gaussian filtering can obscure individual differences by smoothing across functional areas.
• Temporal non-local means (tNLM) filtering offers noise reduction while preserving spatial structures, but requires careful parameter selection.

Purpose of the Study:

• To introduce and evaluate a novel data-dependent filtering method for resting fMRI data.
• To improve noise reduction in fMRI while preserving critical spatial information and individual differences.
• To overcome the limitations of traditional filtering techniques in functional brain connectivity analysis.

Main Methods:

• Development of a Global PDF-based tNLM (GPDF) filtering technique.
• Utilizing a data-dependent optimized kernel function for tNLM filtering.
• Applying global filtering to enhance noise reduction without spatial blurring.

Main Results:

• GPDF filtering demonstrates improved noise reduction capabilities in fMRI data.
• The method effectively preserves spatial structures and functional boundaries.
• Optimized kernel selection leads to superior performance compared to standard methods.

Conclusions:

• GPDF filtering offers a significant advancement for resting fMRI analysis.
• This method enhances the characterization of functional brain connectivity by improving data quality.
• The technique holds promise for more accurate and reliable neuroimaging studies.