Age and Sex-Related Effects on Single-Subject Gray Matter Networks in Healthy Participants

Abstract

Recent developments in image analysis have enabled an individual’s brain network to be evaluated and brain age to be predicted from gray matter images. Our study aimed to investigate the effects of age and sex on single-subject gray matter networks using a large sample of healthy participants. We recruited 812 healthy individuals (59.3 ± 14.0 years, 407 females, and 405 males) who underwent three-dimensional T1-weighted magnetic resonance imaging. Similarity-based gray matter networks were constructed, and the following network properties were calculated: normalized clustering, normalized path length, and small-world coefficients. The predicted brain age was computed using a support-vector regression model. We evaluated the network alterations related to age and sex. Additionally, we examined the correlations between the network properties and predicted brain age and compared them with the correlations between the network properties and chronological age. The brain network retained efficient small-world properties regardless of age; however, reduced small-world properties were observed with advancing age. Although women exhibited higher network properties than men and similar age-related network declines as men in the subjects aged < 70 years, faster age-related network declines were observed in women, leading to no differences in sex among the participants aged ≥ 70 years. Brain age correlated well with network properties compared to chronological age in participants aged ≥ 70 years. Although the brain network retained small-world properties, it moved towards randomized networks with aging. Faster age-related network disruptions in women were observed than in men among the elderly. Our findings provide new insights into network alterations underlying aging.

Single‐subject cortical morphological brain networks across the adult lifespan

“…Thus, our findings indicate continually decreased functional segregation of the CT networks during the age‐related process. This is consistent with previous studies of single‐subject morphological brain networks based on gray matter volume, which found a negative correlation for clustering coefficient and local efficiency with age (Kong et al, 2015 ; Shigemoto et al, 2023 ). However, using the same method as this work, a recent study found that clustering coefficient of CT networks showed an inverted U‐shaped age‐related trajectory (Wang et al, 2022 ).…”

“…Thus, our results suggest a fluctuant change in functional integration of the CT networks across the adult lifespan: decreasing between ages 18 and 41, then increasing between ages 41 and 57, and finally decreasing until age 88. Notably, the cubic age‐related changes are first reported since none of previous studies utilized the cubic model to examine the age‐related changes in single‐subject morphological brain networks (Kong et al, 2015 ; Shigemoto et al, 2023 ; Wang et al, 2022 ). Moreover, findings from the previous studies were different or even opposite possibly due to different statistical models and network construction methods.…”

“…Using the newly developed approaches, two previous studies have explored early cortical development of the neonatal brain (Fenchel et al, 2020 ; Galdi et al, 2020 ). More recently, two large sample studies further examined single‐subject morphological brain networks across the adult lifespan (Shigemoto et al, 2023 ; Wang et al, 2022 ). However, these studies constructed single‐subject morphological brain networks either based on a single morphological feature (i.e., gray matter volume) or by integrating multiple features into a single model.…”

Frontal deficits and atrophy in a patient with familial encephalopathy with neuroserpin inclusion bodies detected by single-case voxel-based morphometry: a case report

“…To predict the patient’s brain age, the support regression model implemented in the LIBSVM ( http://www.csite.ntu.edu.tw/~cjlin/libsvm/ ) toolbox with a linear kernel and default set of parameters was used (i.e., in the LIBSVM: C = 1, v = 0.5). The details of the analytical method were described in a previous study [ 13 ]. For the regression model, the chronological age was considered the dependent variable, whereas the principal components derived from the gray matter voxel intensities were considered independent variables.…”

Frontal deficits and atrophy in a patient with neuroserpinosis detected by single-case voxel-based morphometry: A case report

“…To predict the patient's brain age, the support regression model implemented in the LIBSVM (http://www.csite.ntu.edu.tw/~cjlin/libsvm/) toolbox with a linear kernel and default set of parameters was used (i.e., in the LIBSVM: C = 1, v = 0.5). The details of the analytical method were described in a previous study 12 . For the regression model, the chronological age was considered the dependent variable, whereas the principal components derived from the gray matter voxel intensities were considered independent variables.…”