The Voxel-Wise Functional Connectome Can Be Efficiently Derived from Co-activations in a Sparse Spatio-Temporal...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Network Science

Background:

• Mapping the human brain's functional connectome is crucial for understanding neurological function and disease.
• Current methods using functional Magnetic Resonance Imaging (fMRI) are computationally intensive and require large datasets.
• A deeper understanding of resting-state functional connectivity is needed to refine these mapping efforts.

Purpose of the Study:

• To develop a computationally efficient method for estimating functional connectivity from fMRI data.
• To validate a novel sparse representation of fMRI signals for network analysis.
• To demonstrate the utility of this method in studying brain changes during sleep and aging.

Main Methods:

• Introduced a sparse representation of fMRI data using a discrete point-process encoding of high-amplitude blood oxygenation level-dependent (BOLD) signal events.
• Validated the method by comparing results with standard voxel-wise linear correlation matrices on two independent datasets (n=71 and n=1147).
• Assessed changes in node strength during deep sleep and age-related network reorganization.

Main Results:

• The point-process method accurately estimated functional connectivity, replicating findings from standard correlation techniques.
• This novel approach requires only ~1% of the original BOLD signal data, offering significant dimensionality reduction.
• Classification accuracy was maintained or improved compared to standard linear correlations, demonstrating the method's robustness.

Conclusions:

• A sparse representation of fMRI data, focusing on high-amplitude BOLD signal events, is sufficient for comprehensive functional connectome mapping.
• This dimensionality reduction technique offers a computationally efficient alternative for analyzing large-scale neuroimaging datasets.
• The findings suggest that underlying electrophysiological signals may manifest as temporally localized, all-or-none events, akin to neural avalanches.