Evaluating the reliability of different preprocessing steps to estimate graph theoretical measures in resting state...

Area of Science:

• Neuroimaging
• Computational Neuroscience
• Brain Connectomics

Background:

• Resting-state functional MRI (rs-fMRI) enables human brain connectome quantification.
• Numerous preprocessing and post-processing methods exist, influencing connectivity results.
• The impact of different preprocessing pipelines on reliability and reproducibility is not fully understood.

Purpose of the Study:

• To assess the reliability and reproducibility of graph theoretical measures derived from rs-fMRI data.
• To evaluate seven distinct preprocessing schemes for their impact on brain connectivity analysis.
• To investigate the influence of head motion on graph theoretical measurements across different preprocessing strategies.

Main Methods:

• rs-fMRI data from healthy controls were analyzed using seven different preprocessing schemes.
• The brain was parcellated into 190 nodes.
• Four graph theoretical measures were calculated: global efficiency (GEFF), characteristic path length (CPL), average clustering coefficient (ACC), and average local efficiency (ALE).

Main Results:

• Significant differences in graph theoretical measures were observed based on the selected preprocessing steps.
• Head motion demonstrated a dependency with graph theoretical measurements in most preprocessing strategies.
• Implementing outlier censoring in the functional time-series processing reduced motion dependency and increased measurement reliability.

Conclusions:

• The choice of preprocessing steps critically affects rs-fMRI derived brain connectome metrics.
• Head motion is a significant confound in graph theoretical analyses of brain connectivity.
• Censoring functional time-series outliers enhances the reliability of graph theoretical measurements and mitigates motion-related artifacts.