Reproducibility of graph measures derived from resting-state MEG functional connectivity metrics in sensor and source...

Area of Science:

• Neuroscience
• Network Science
• Biomarker Discovery

Background:

• Graph analysis of resting-state magnetoencephalography (MEG) is used to study brain networks in neurological disorders.
• Lack of standardized network construction strategies hinders reproducibility and clinical application of graph measures.
• Ensuring global graph measures are consistent across sensor and source space analyses is crucial for biomarker development.

Purpose of the Study:

• To investigate the test-retest reliability of global graph measures.
• To assess the association between sensor and source space analyses for graph measures.
• To guide the selection of optimal network construction strategies for reproducible biomarker discovery using MEG.

Main Methods:

• Utilized resting-state MEG data from 89 healthy subjects (Human Connectome Project).
• Employed atlas-based beamforming for source reconstruction.
• Estimated functional connectivity (FC) using debiased weighted phase lag index (dwPLI) and amplitude envelope correlation (AEC) metrics across six frequency bands in both sensor and source spaces.
• Examined reliability and sensor-source association across varying network densities using proportional weight and orthogonal minimum spanning tree thresholding.

Main Results:

• Graph measures demonstrated fair to excellent test-retest reliability and significant sensor-source association across most FC metrics and frequency bands at 100% network density.
• Amplitude metrics (AEC) yielded the highest reliability and sensor-source consistency.
• Reliability was comparable between sensor and source spaces for amplitude metrics, while dwPLI showed higher reliability in source space for higher frequencies.

Conclusions:

• Graph measures derived from MEG are reliable and reproducible, supporting their use as potential clinical biomarkers for neurological disorders.
• Amplitude-based functional connectivity metrics are recommended for constructing brain networks due to their superior reliability and sensor-source consistency.
• The findings provide crucial guidance for selecting appropriate network construction strategies in MEG research, particularly for amplitude synchrony investigations.