The (in)stability of functional brain network measures across thresholds

Area of Science:

• Neuroscience
• Network Science
• Graph Theory

Background:

• Brain organization exhibits complex network properties quantifiable via graph theory.
• Functional brain organization is typically derived from continuous correlation data, necessitating thresholding for binary graph analysis.
• Current lack of consensus on thresholding methods leads to potential instability in network measures and group comparisons.

Purpose of the Study:

• To evaluate the stability of brain network measures across various thresholds using resting-state functional connectivity data.
• To compare the stability of network measures between absolute and proportional thresholding strategies.
• To investigate the impact of threshold choice on group differences (sex and age).

Main Methods:

• Analysis of a large resting-state functional connectivity dataset.
• Calculation of network measures across absolute (correlation-based) and proportional (sparsity-based) thresholds.
• Comparison of network measure stability and group differences across different thresholding methods.

Main Results:

• Network measures demonstrated significant instability across absolute thresholds, with potential reversals in group difference directions.
• Network measures exhibited greater stability when analyzed using proportional thresholds.
• Results highlight threshold-dependent variations in network properties and group comparisons.

Conclusions:

• Thresholding functional connectivity data requires careful consideration due to the instability of network measures, particularly with absolute thresholds.
• Proportional thresholding offers a more stable approach for analyzing brain networks.
• Researchers should exercise caution when interpreting findings from binary graph models derived from functional connectivity data.