Abstract
The brain is a dynamic and complex network formed by hierarchical interconnections among functional regions and neural systems across lifespan. Recent studies using graph theory explore the functional organization of resting-state brains have coherently indicated many non-random topological properties such as modular structure, central communication hubs, small-world property, scale-free network, and efficient network architecture. However these studies have focused mainly on partial brain attributes rather than integrated communication across brain regions and the organization or order of the brain. In this study, we explored the entire lifespan trajectory of human whole-brain organization used resting-state functional magnetic resonance imaging (rs-fMRI) and entropy analyses of graphic network. We found that the network eigen entropy (NEE) decreased from child to early middle aged but remained stable from middle aged to early old. Further, the slightly increased NEE was observed in the older people. Our findings not only revealed the U-shaped development and aging trajectory of functional organization in resting-state human that underpin variance in cognitive functions of brains across lifespan.
Introduction
The human brain is an intricate, dynamic network. While graph theory has successfully identified local attributes like "small-world" properties or specific hub regions, these metrics often fail to capture the global state of the brain's organization—its overall "order" or "disorder." This study introduces Network Eigen Entropy (NEE) as a robust quantitative index to map the trajectory of whole-brain organization across the human lifespan, from childhood through old age.
Methods
The researchers analyzed resting-state functional magnetic resonance imaging (rs-fMRI) data from the NKI / Rockland Sample (NKI-RS), which included 173 healthy participants aged 7 to 85 years.
To calculate NEE, the brain was first partitioned into nodes at three different spatial scales (160, 500, and 808 regions) to ensure findings were scale-invariant. The importance of each node was determined using Eigenvector Centrality (EC), which weights connections based on the influence of the connected neighbors. Finally, Shannon entropy was applied to the distribution of these EC values to derive the NEE. A high NEE indicates a more random or disordered system, while a low NEE signifies a more structured and organized network. General Linear Models (GLM) were used to control for head motion and sex, and the Akaike Information Criterion (AIC) helped identify the best-fitting mathematical model for the aging process.
Results
The study revealed a distinct U-shaped trajectory for brain entropy across the lifespan. NEE values were highest in childhood, dropped significantly through early adulthood, remained stable during middle age, and showed a slight increase in the elderly.
Statistical analysis confirmed that a quadratic model provided a significantly better fit than a linear one across all spatial scales. For the 808-node scale, the quadratic fit achieved an $R^2 = 0.42$ ($p < 0.001$). This trend was scale-invariant, appearing consistently regardless of how the brain was partitioned. Specific regions, such as the Left Supplementary Motor Area (SMA.L) and the SupraMarginal Gyrus, were identified as primary drivers of these organizational shifts.
Discussion
The high NEE observed in childhood suggests a state of "productive disorder," where a flexible, less specialized network facilitates diverse learning. As the brain matures, it "prunes" connections, leading to the highly organized and efficient state seen in middle age. The subsequent increase in NEE in older adults likely reflects a loss of this precise organization, potentially explaining age-related cognitive decline as functional connectivity becomes less precise.
Limitations and Future Work
This study was cross-sectional; longitudinal data following the same individuals over time would further validate these findings. Additionally, while the authors link NEE to cognitive function, direct correlations with behavioral test scores were not performed in this specific analysis. Future research could also investigate how pathological conditions like Alzheimer's disease alter this entropy trajectory compared to healthy aging.
Summary updated