From nodes to networks: How methods for defining nodes influence inferences regarding network interactions

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Brain Connectivity Analysis

Background:

• Functional connectivity (FC) analysis in fMRI relies on identifying network nodes from activation patterns.
• Existing methods like Activation Likelihood Estimate (ALE) identify canonical activation maxima.
• Experimentally Derived Estimates (EDE) use individual participant data for activation maxima.

Purpose of the Study:

• To compare ALE and EDE approaches for network node identification in working memory tasks.
• To evaluate their effectiveness in functional inference and connectivity analysis.
• To determine the optimal method for individualized network inference.

Main Methods:

• Localized network nodes using both ALE and EDE for each participant in verbal and visual working memory tasks.
• Extracted time series from identified nodes.
• Compared nodal time series and network edge correlations between ALE and EDE methods.

Main Results:

• Significant differences were observed in a large proportion of network edge correlations between ALE and EDE methods.
• ALE effectively summarizes working memory network node locations across studies and subjects.
• EDE methods demonstrated sensitivity to individual functional loci, providing individualized connectivity estimates.

Conclusions:

• ALE provides a generalized summary of working memory network nodes.
• EDE offers superior individualized estimates of functional connectivity.
• A hybrid approach combining ALE and EDE is proposed as optimal for robust network inference.