Spatial Smoothing Effect on Group-Level Functional Connectivity during Resting and Task-Based fMRI.

Area of Science:

• Neuroimaging analysis
• Brain network connectivity

Background:

• Spatial smoothing is crucial for neuroimaging data quality.
• Kernel size selection is challenging, potentially altering image and network properties.
• Limited data exists on Gaussian kernel size effects on group-level neuroimaging results.

Purpose of the Study:

• Investigate the influence of Gaussian kernel size on functional connectivity networks.
• Evaluate effects on network parameters and principal component analysis (PCA)/independent component analysis (ICA) metrics.
• Assess these impacts in whole-brain resting-state (rs-fMRI) and task-based (tb-fMRI) data from healthy adults.

Main Methods:

• Applied Gaussian kernels of {0, 2, 4, 6, 8, 10} mm to rs-fMRI and tb-fMRI data.
• Analyzed graph theoretical measures: betweenness centrality, global/local efficiency, clustering coefficient, average path length.
• Evaluated PCA and ICA parameters (kurtosis, skewness) for functional images across kernel sizes.

Main Results:

• Kernel size significantly alters node connections and functional network structures.
• Modifications were observed in PCA and ICA parameters based on kernel size.
• Graph theoretical network metrics demonstrated relative stability across different kernel sizes.

Conclusions:

• Gaussian kernel size is a critical parameter influencing neuroimaging data preprocessing.
• While network structure and certain analysis parameters are sensitive to smoothing, key graph metrics remain robust.
• Findings provide guidance for optimizing spatial smoothing in fMRI studies.