Age‐related Modularity Changes of the Structural Brain Network: The Framingham Study
Stephan Seiler, Evan Fletcher, Alexa S. Beiser, Jayandra Jung Himali, Claudia L. Satizabal, Sudha Seshadri, Pauline Maillard, Charles Decarli- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
While studies have consistently shown that brain network efficiency decreases with age, data on age‐related reconfiguration of network modules are scarce. We assessed structural brain networks from healthy adults of the Framingham Study to explore age‐related modularity differences.
Method
Structural brain networks of 2,903 healthy adults (mean age 55, SD 0.96, range 25‐95, 53% female) were obtained from DTI probabilistic tractography. Connectivity matrices were constructed from 72 cortical and subcortical regions. First, we calculated efficiency to check the validity of our data. Second, we constructed a multilayer network with each layer representing one age decile and calculated modularity for each layer using a robust implementation of the Louvain community‐detection algorithm (Blondel et al., 2008). Third, we compared modular structures between each layer and a reference group (youngest) using the variation of information (VI), an index measure of distance between two modular patterns (e.g. modularity at different ages). Finally, we calculated the participation coefficient, a measure of modular node integration, and the node flexibility, a measure of module allegiance change and assessed their relationships with age.
Result
Efficiency decreased with age (b = ‐0.34, SE 0.02, p<0.001), being stable until approximately 50 and declining linearly thereafter (see Figure 1, panel a). Modularity showed a u‐shaped trajectory with most integration at the fifth decile of age (see Figure1, panel b). VI index revealed that modular reconfiguration was greatest between the youngest and the oldest age decile (0.43). Participation coefficient was highest at the fifth decile (see Figure2, panel c) and in frontal nodes (see Figure2, panel a), where it also exhibited high variability. Frontal variability followed a complex pattern (Figure2, panel b). Conversely, central and temporal‐occipital nodes showed less participation and variability (see Figure2, panels a and b), and the lowest flexibility (see Figure3).
Conclusion
Brain network modules reconfigure with aging. This reconfiguration is non‐uniform. While midline‐near modules, including the cingulate and parieto‐occipital cortices, show relative stability, frontal modules seemed to reconfigure considerably, as seen by increased average flexibility (change in modular allegiance) and less integration with advancing age, as seen by declining participation coefficients.