Abstract Body:

Background: The intricate process of brain aging encompasses a variety of functional shifts within numerous neural subnetworks. Despite this complexity, the majority of prior research on age-related changes in functional connectivity (FC), utilizing resting-state functional magnetic resonance imaging (rs-fMRI), has concentrated on linear correlations, neglecting the nonlinear dynamics inherent in fMRI signal properties. The impact of brain aging on nonlinear causal relationships within extensive neural networks, especially in terms of sex differences, remains understudied. A novel approach to understanding the aging brain may lie in examining alterations in network interaction patterns through the lens of nonlinear causality, a perspective that could elucidate previously identified sex differences in fMRI studies. This approach may also shed light on why certain brain regions or subnetworks are more susceptible to aging effects and provide evidence for the underlying nonlinear mechanisms.

Methods: Our study comprised 227 healthy individuals from the “Leipzig Study for Mind-Body-Emotion Interactions” (LEMON), categorized into two age groups: a younger group aged 20–35 years (N=153, mean age 25.1±3.1 years, 45 females) and an older group aged 59–77 years (N=74, mean age 67.6±4.7 years, 37 females). Participants underwent neuroimaging and cognitive testing, administered by trained undergraduate psychology students following a standardized protocol. Data preprocessing was performed using the Data Processing & Analysis for Brain Imaging (DPABI) software. Nonlinear Granger causality (NGC) was assessed using the neural Granger causality technique at regional and subnetwork levels, and a NGC-based FC matrix was constructed for brain-region-scale rs-fMRI signals with a component-wise long-short term memory network (cLSTM).

Results: The study revealed that brain aging is associated with extensive reductions in NGC at both regional and subnetwork levels, with high reproducibility across varying network densities, thus validating the effectiveness of both static and dynamic analytical methods. Notably, females showed greater variability and decreased stability in NGC with age, particularly in the connectivity between the visual network and other subnetworks. Moreover, the strength of NGC was found to correlate well with individual cognitive function, suggesting its potential as a sensitive indicator in cognitive studies for understanding mechanisms at individual or group levels.

Conclusions: The results underscore the utility of NGC analysis in uncovering sex-specific patterns of brain aging. The pronounced decrease in multi-scale nonlinear interactions appears to be a hallmark of the human brain’s aging process.

Author