Measuring functional connectivity in frequency-domain helps to better characterize brain function

Area of Science:

• Neuroimaging
• Brain Connectivity
• Functional Magnetic Resonance Imaging (fMRI)

Background:

• Resting-state functional connectivity (FC) is a key tool in fMRI for brain region analysis, phenotype prediction, and disease diagnosis.
• Current FC analysis primarily focuses on temporal correlations, with limited exploration of frequency-domain interactions.
• Understanding functional interactions across different frequencies is crucial for a comprehensive brain analysis.

Purpose of the Study:

• To investigate the utility of frequency-domain analysis, specifically coherence, for measuring functional connectivity in fMRI data.
• To compare the performance of coherence-based FC with traditional correlation-based FC in machine learning tasks.
• To determine if frequency-domain analysis offers advantages in identifying brain functional borders and improving pattern discrimination.

Main Methods:

• Functional connectivity (FC) was assessed using both coherence (frequency-domain) and correlation (time-domain) methods.
• Two machine learning tasks were employed to evaluate the pattern discrimination capabilities of each FC approach.
• fMRI data was analyzed to contrast the effectiveness of temporal correlation versus frequency-domain coherence.

Main Results:

• Coherence-based functional connectivity (FC) demonstrated superior performance in identifying finer functional subregions within the brain.
• Frequency-domain analysis using coherence led to enhanced pattern discrimination capabilities compared to temporal correlation.
• The study confirmed the feasibility and added value of using coherence for fMRI analysis.

Conclusions:

• Modeling functional interactions in the frequency domain, particularly using coherence, provides richer information than time-domain analysis for fMRI.
• Frequency-domain FC analysis offers a promising new perspective for advancing functional neuroimaging and brain mapping.
• This approach can potentially improve the accuracy of identifying brain functional borders and diagnosing neurological and psychiatric conditions.