Age-related degradation of cerebellar functional network topology

Abstract

With the aging of global populations, understanding the role of the cerebellum in aging is crucial. However, existing studies primarily focus on structural and functional changes, leaving the topological properties of cerebellar functional networks in aging unclear. Here, we explored the aging effects on the topological organization of cerebellar functional networks and investigated the relationship between these changes and cognitive function. The results revealed that, compared with young adults, older adults exhibited declines in the global and local network efficiency, modularity, and small-world properties of the cerebellum. Additionally, reductions were observed in the betweenness centrality within the visual network (VN), the nodal efficiency (NE) within the VN and lateral somatomotor network (SMN(Lat)). Moreover, the nodal local efficiency (NLE) decreased in the default mode network and SMN(Lat) of older adults. Notably, the gamma value of the cerebellar network was positively correlated with Stroop task accuracy in the incongruent condition among older adults. Additionally, aging impaired whole-brain (including cerebellar) network efficiency but preserved modularity and small-world properties, and at the nodal level, cerebellar regions showed selective reductions in NE and NLE, alongside degree centrality reorganization. Exploratory analyses revealed gender-specific patterns in cerebellar network topology during aging, with males experiencing relatively limited declines and females showing more extensive deterioration. The findings indicated that aging is associated with disruptions in the global and regional topology of the cerebellar functional connectome, with females exhibiting more pronounced deterioration than males. The integrity of cerebellar network topology appeared crucial for maintaining cognitive control.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-02059-w.

Highlights

• Graph-theoretical analysis reveals distinct global and regional alterations in brain connectivity among young and old adults.

• Older adults exhibit significant declines in the global and local efficiency, modularity, and small-world characteristics of cerebellar functional networks.

• Aging is associated with reduced cerebellar betweenness centrality in the visual network, impaired nodal efficiency in the visual and somatomotor networks, and diminished nodal local efficiency in the default mode and somatomotor networks within critical brain networks.

• A potential relationship is identified between cerebellar network topological properties and cognitive control function during aging.

• Exploratory analyses reveal gender differences in cerebellar network topology during aging, with males experiencing limited declines and females showing more extensive deterioration.

• These findings highlight the potential of cerebellar functional connectome analysis for the early detection of aging.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1007/s11357-025-02059-w.

Introduction

For a considerable period, the brain has remained at the core of neuroscience research. Nevertheless, the cerebellum, a crucial component of the brain, is frequently overlooked. This neglect is impeding progress in neuroscience [1]. Studies have revealed that the cerebellum, accounting for just 10.3% of the brain’s total mass, harbors as much as 80.2% of the brain’s neurons—more than four times the number found in the cerebrum [2]. From an evolutionary perspective, the cerebral cortex and cerebellum exhibit coordinated scaling of neuronal numbers: for each neuron added to the cerebral cortex, four neurons are generated in the cerebellum [3]. Beyond its remarkable neuron count, the cerebellum also presents unique structural characteristics. The human cerebellum is more densely folded than the cerebral cortex, with its cortex encompassing approximately 80% of the total surface area of the cerebral cortex [4]. The cerebellum and cerebral cortex increase in mass simultaneously [3, 5]. Functionally, the majority of the cerebellum exhibits functional coupling with cerebral association networks, with half of its connections targeting the cerebellar networks related to cognitive control and the default network, both of which play significant roles in higher-level cognition [6–8]. These findings underscore the cerebellum’s critical role in brain function.

As the global population ages rapidly, comprehending the impacts of brain aging on older adults has become increasingly crucial. Research indicates that the cerebellum exhibits signs of senescence at an earlier stage compared to the hippocampus [9]. However, only a few studies have explored the cerebellum’s role in aging [10]. Existing research has primarily focused on structural changes and functional abnormalities in the cerebellar regions during aging, revealing a correlation between changes in the cerebellar and cognitive decline. During the aging process, the white matter (WM) and gray matter (GM) in the cerebellum exhibit distinct trajectories. Specifically, WM volume follows an inverted U-shaped pattern, peaking around age 10 and slowly increasing until ages 30–40 before declining [11]. In contrast, GM demonstrates rapid growth during the first 10 years of life, followed by a steady decrease thereafter [11]. Moreover, specific regions of the cerebellum also show a close association with cognitive function during aging. One study found that larger left Crus I volume was tendentially associated with better spatial working memory performance, while larger posterior cerebellum volume and increasing age were linked to worse performance [12]. Intra-cerebellar connectivity also declines with age, particularly in connections involving lobules IV, IX, VIII A, and VIII B, which may lead to poor cognitive performance [13]. Additionally, in cerebr-cerebellar connectivity, young adults typically have more localized and functionally specific cortico-cerebellar networks, while older adults show more widespread connectivity patterns [14]. However, the brain is an extremely complex and sophisticated information-processing system. To comprehensively understand the role of the cerebellum in aging, it is essential to explore the topological properties of cerebellar functional networks.

Graph theoretical analysis has emerged as a crucial tool in network neuroscience [15]. In graph theory, networks are conceptualized as sets of nodes connected by edges, where nodes represent cerebral regions and edges indicate neural connections [16]. It provides a framework to quantify the connectivity of the human brain as a topological network at global, modular, and nodal levels [15, 17]. This approach is crucial for understanding the architecture, development, and evolution of brain networks [18]. Not only does the brain network possess these remarkable characteristics, but the cerebellar functional connectome also shows similar and unique features. Research has shown that the brain’s networks exhibit small-world properties (high clustering and high global efficiency), a modular community structure, and a heavy-tailed degree distribution (with highly connected hub nodes or hubs) [19]. Analogously, the cerebellar functional connectome demonstrates integrated small-world organization, functionally segregated modules (attention/executive network, default mode network (DMN) and task-positive network), and multi-scale hierarchical organization, which support the cerebellum’s role in motor control and higher-order cognitive functions [20, 21]. However, changes in topological properties occur in the whole-brain functional connectome throughout nearly the entire human lifespan [22]. Research has revealed that age impairs the economic performance of small-world brain networks [23]. Specifically, older adults exhibit reduced global and local efficiency, as well as maximum-cost efficiency, in brain functional networks. This reduction not only implies a weakened ability to process information but also indicates a diminished capacity to strike a balance between information processing efficiency and cost. Furthermore, the brain’s functional network shows a linear decrease in modularity with age, which reflects a reduction in functional network differentiation [22, 24]. Additionally, the number of hubs diminishes with age, and their spatial distribution shifts toward the frontoparietal network (FPN) and some higher-order cognitive networks [25]. The collective changes in network measures are significantly correlated with cognitive performance [26]. Although previous studies have provided important insights, the effects of aging on cerebellar topological organization and their association with cognitive function remain unclear. Furthermore, integrating the cerebellum into whole-brain analysis is essential to delineate its unique contribution to the aging process.

Therefore, this study aims to explore the aging effects on the topological organization of cerebellar functional networks and to further investigate the relationship between these changes and cognitive function. Furthermore, the cerebellum is integrated into a whole-brain analysis framework to examine its topological properties. By employing graph theoretical analysis, we hope to provide novel insights into the neural substrates of age-related cognitive decline and contribute to a more comprehensive understanding of the cerebellum’s role in healthy aging. Additionally, exploratory analyses are performed by separately comparing young and older participants within males and females.

Methods

Participants

The present study included 89 participants, consisting of 54 young adults (age range 20–35 years, mean age 25.04 ± 3.87 years, 21 females) and 35 older adults (age range 61–76 years, mean age 67.80 ± 4.14 years, 25 females). All data were obtained from an ongoing database that investigates age-related cognitive control. The research received approval from the Institutional Review Board at the Brain Imaging Center of Southwest University and adhered to the ethical guidelines of the Declaration of Helsinki.

The eligibility criteria for participants were as follows: all participants were required to be right-handed and native Chinese speakers. Older adults were required to score above 26 on the Mini-Mental State Examination (MMSE) to ensure normal cognitive function [27]. Individuals were excluded if they had a history of neuropsychiatric disorders, head trauma, substance abuse, vision abnormalities (including corrected/uncorrected impairment or color blindness), fewer than 6 years of education, or head motion exceeding 3.0 mm or 3.0° during functional magnetic resonance imaging (fMRI) acquisition. Written informed consent was obtained from all participants before their involvement in the study.

Methodological overview

Resting-state fMRI (rs-fMRI) is a neuroimaging technique that captures intrinsic, spontaneous blood oxygen level-dependent (BOLD) signals from the resting brain without requiring participants to perform any cognitive tasks [28, 29]. This method provides a powerful approach for investigating the topological organization of brain networks and serves as the fundamental technical basis for constructing cerebellar and whole-brain functional networks in this study [29].

Images data acquisition

All magnetic resonance imaging (MRI) data were acquired using a 3.0 T Siemens Prisma scanner equipped with a 12-channel head coil. Head motion was minimized using comfortable foam padding, and scanner noise was reduced with earplugs.

Rs-fMRI data were obtained using an echo-planar imaging (EPI) sequence. Visual stimuli were presented using E-Prime 2.0 software on an 18.5-inch monitor (resolution of 1366 × 768 at 60 Hz). Participants viewed the display via a mirror mounted on the head coil. The scanning protocol consisted of two phases: 7.5 min with eyes open followed by 7.5 min with eyes closed. Participants were instructed to remain relaxed throughout the entire process and to avoid falling asleep during the eyes-closed phase. The scanning parameters for the EPI scans were as follows: repetition time (TR) = 2000 ms; echo time (TE) = 30 ms; flip angle = 90^°; field of view (FOV) = 224 × 224 mm²; 62 slices; voxel size = 2 × 2 × 2.3 mm³; 450 volumes (Fig. 1A).

Fig. 1.

Workflow and analytical sequence for the present research. A. Participants, cognitive assessment, and image acquisition. B. Cerebellar network construction and analysis. C. Statistical analysis methods. Abbreviations: M: male; F: female; MMSE: Mini-Mental State Examination; DAN, dorsal attention network; DMN, default mode network; VN, visual network; FPN, frontoparietal network; SN, salience network; CON, cingulo-opercular network; SMN(Dor), somatomotor network-dorsal; SMN(Lat), somatomotor network-lateral

High-resolution T1-weighted structural images were also obtained for each subject using a high-resolution, magnetization-prepared rapid gradient-echo (MPRAGE) sequence with the following parameters: TR = 2530 ms; TE = 2.98 ms; flip angle = 7^°; resolution matrix = 256 × 256; slice thickness = 1.0 mm; voxel size = 1 × 1 × 1 mm³; 192 contiguous slices (Fig. 1A).

Image preprocessing

Image preprocessing was performed using the SPM12 and DPABI toolbox in Matlab 2019b [30]. The preprocessing pipeline included the following steps: (1) removal of the first ten images from each participant to stabilize the magnetic field and ensure participant adaptation; (2) spatial realignment to account for head movements, with participants exhibiting head motion exceeding 3 mm or 3 degrees being excluded; (3) co-registration between the mean functional image and the corresponding structural image; (4) regression of nuisance variables, including the Friston-24 motion parameters and two other confounding signals (WM, cerebrospinal fluid); (5) spatial normalization to the Montreal Neurological Institute (MNI) template and resampling to a resolution of 3 × 3 × 3 mm³; (6) temporal band-pass filtering within the frequency range of 0.01–0.08 Hz; and (7) spatial smoothing using a Gaussian kernel with a full width at half maximum (FWHM) of 6 mm.

Cerebellar functional network construction

To explore topological differences between young and old groups, weighted undirected graph metrics were employed, as this approach retains greater biologically relevant information compared to binary undirected graph metrics [31]. Network construction utilized 27 cerebellar nodes from the Seitzman atlas, which organizes the cerebellum into 8 functional networks [32]. Each node was represented by a 4-mm spherical region of interest (ROI) centered on standardized coordinates, as detailed in Supplementary Information 1. The average time series extracted from each spherical region formed the basis for connectivity analysis. For each participant, a 27 × 27 correlation matrix was constructed by calculating the Pearson correlation coefficients (PCC) between the time series of all pairs of brain regions. To further analyze these matrices, a Fisher r-to-z transformation was applied. Weighted topological parameters were computed for each subject’s positive functional connections across a sparsity range of 0.10–0.34, with an interval of 0.01 [33]. The area under the curve (AUC) across this sparsity range was calculated for the cerebellar functional network measures. The AUC was chosen for statistical analysis due to its enhanced sensitivity and ability to provide a summarized scalar for topological characterization, which is independent of single threshold selection [33, 34] (Fig. 1B).

Whole-brain functional network construction

Consistent with the construction of the cerebellar functional network, 300 whole-brain nodes (including cerebellar nodes) from the Seitzman atlas were utilized [32]. These nodes are categorized into 13 functional networks, including an unlabeled category [32]. According to the Seitzman atlas, cerebellar nodes were defined as 4-mm spherical ROIs centered on standardized coordinates, while all other nodes were defined as 5-mm spherical ROIs [32]. The average time series derived from each spherical ROI served as the foundation for connectivity analysis. For each participant, a 300 × 300 correlation matrix was generated by computing the PCC between the time series of all pairs of brain regions. Subsequently, a Fisher r-to-z transformation was applied to these matrices to facilitate further analysis. Weighted topological parameters were calculated for the positive functional connections of each subject across a sparsity range of 0.10 to 0.34 (0.01 interval) [33]. For each topological measurement, the AUC was calculated across this sparsity range.

Network analysis

The topological properties of cerebellar and whole-brain functional connectomes were computed at global and nodal levels across all sparsity thresholds using the Gretna toolbox [35].

The calculated global properties included the following: (1) Network efficiency, which consists of global efficiency (E_glob) and local efficiency (E_loc). E_glob represents the capacity of a network to transfer information across the whole network, whereas E_loc signifies the network’s capability to convey information at the local level [23]. (2) Modularity (Q) quantifies the extent to which a network is organized into distinct modules or communities, with higher values of Q indicating more pronounced modular organization [36]. (3) Small-world metrics include clustering coefficient (C_p), characteristic path length (L_p), normalized clustering coefficient (γ, gamma), normalized characteristic path length (λ, lambda), and small-worldness (σ, sigma). Specifically, C_p quantifies the proportion of a node’s neighbors that are also neighbors of one another; L_p represents the mean distance between all possible pairs of nodes within the network; γ and λ compare C_p and L_p, respectively, between the real network and random networks. σ is defined as the ratio γ/λ. The cerebellar functional networks demonstrate characteristic small-world properties when γ > 1 and λ ≈ 1, or alternatively when σ > 1 [37].

Additionally, nodal centrality metrics, including degree centrality (DC), betweenness centrality (BC), nodal efficiency (NE), and nodal local efficiency (NLE), were calculated for each node in the network. DC quantifies the number of connections attached to a node, which reflects the node’s capacity for information communication within the network; BC evaluates the proportion of all shortest paths in the network that traverse a given node, effectively depicting its impact on the information flow among other nodes; NE is derived from the reciprocal of the average of the shortest path lengths between the node and all other nodes, serving as an indicator of the parallel information transfer efficiency of the network nodes; and NLE refers to the communication efficiency among the node’s neighbors after the node has been removed [38, 39]. These parameters enable the identification of key hub nodes that facilitate information flow across the cerebellar network.

Cognitive assessment

MMSE

Cognitive performance was evaluated via the MMSE in older adults, encompassing five domains: orientation, immediate memory, calculation, recall, and language ability. Scores range from 0 to 30, with higher scores indicating better cognitive function [40].

Stroop task

This study utilized the Stroop task to assess participants’ cognitive control abilities, with the experiment conducted using E-Prime 2.0 [41, 42]. The Stroop effect takes place when it takes longer to name the font color of a word if the color contradicts the semantic meaning of the word. This is because of the interference between language and color processing, and resolving such interference depends on strong cognitive control [43]. A mixed factorial design was employed, featuring a 2 (congruency condition: congruent vs. incongruent) × 2 (group: young adults vs. older adults) structure, where congruency condition served as the within-subject factor and group as the between-subject factor. The experimental materials consisted of the Chinese characters for “red”, “green”, “yellow”, and “blue”, which were presented in the colors red (RGB: 255,0,0), green (RGB: 0,255,0), yellow (RGB: 255,255,0), and blue (RGB: 0,0,255), respectively. The experiment was designed in a block format, comprising 8 blocks, each containing 20 trials. The congruent condition referred to the situation where the color of the text matched its semantic meaning, while the incongruent condition was the opposite. The study used pseudo-random presentation of a single stimulus within each block. Each trial began with a 500 ms fixation cross (“ + ”) followed by a 2500 ms stimulus presentation. Participants responded using right-hand key presses: the thumb for red, the index finger for green, the middle finger for yellow, and the ring finger for blue. A 20 s rest period was provided between each block. Four raw scores were selected for statistical analysis: the accuracy rate (ACC) and reaction times (RT) were calculated separately for the congruent and incongruent conditions. Additionally, the Stroop effect value was calculated by subtracting the scores obtained under the congruent condition from those under the incongruent condition.

Statistical analysis

Behavioral data were analyzed using SPSS 21.0 software. Group differences in demographic characteristics were examined using appropriate statistical tests. Specifically, chi-square tests were conducted for categorical variables (e.g., sex), and two-sample t-tests were used for continuous variables (e.g., education level).

Topological properties were compared between young and older participants using two-sample t-tests, with sex and education level controlled as covariates, using the Gretna toolbox [35]. To examine the relationship between topologically significant network metrics and cognitive function, partial correlation analysis (controlling for age, sex, and education level) was conducted. Multiple comparisons of nodal features and partial correlation analyses between nodal metrics and cognitive performance were adjusted using the false discovery rate (FDR) method, with a significance threshold set at p < 0.05.

A repeated measures analysis of variance (ANOVA) was conducted on the RT and ACC of Stroop task performance, with imbalanced demographic variables included as covariates. Additionally, a post-hoc two-sample t-test was used to compare the Stroop effect between young and older adults.

Furthermore, to investigate whether resting-state conditions modulate age-related differences in cerebellar network topological properties, a 2 (age groups: young, older) × 2 (resting-state conditions: eyes open, eyes closed) repeated measures ANOVA was employed, with imbalanced demographic variables included as covariates. To investigate sex-specific age-related differences in cerebellar network topological properties, we conducted additional exploratory analyses by performing separate comparisons between young and older participants within males and females. Moreover, abnormalities at the nodal level were visualized on the brain using BrainNet Viewer [44] (Fig. 1C).

Results

Differences in cerebellar global properties between young and older adults

Regarding network efficiency, the values of E_glob (t = 3.587, p < 0.001) and E_loc (t = 2.530, p = 0.013) were significantly higher in young adults compared to older adults (Fig. 2). Similarly, modularity analysis revealed that young adults exhibited a significantly higher modularity index (Q) (t = 3.952, p < 0.001) (Fig. 3).

Fig. 2.

Differences in cerebellar network efficiency between young and old adults. A. Violin plots were presented to show the AUC parameters of E_glob and E_loc for young and old adults, with means and standard deviations illustrated. *: p < 0.05. **: p < 0.01. ***: p < 0.001. B. E_glob and E_loc values were shown across a density range from 10 to 34%. *: p < 0.05

Fig. 3.

Differences in cerebellar modularity (Q) between young and old adults. A. Violin plots were presented to show the AUC parameters of modularity (Q) for young and old adults, with means and standard deviations illustrated. *: p < 0.05. **: p < 0.01. ***: p < 0.001. B. Modularity (Q) was shown across a density range from 10 to 34%. *: p < 0.05

In terms of small-world metrics, the findings indicated that compared to young adults, older adults exhibited a significantly higher value of L_p (t = −3.583, p < 0.001), while there was no significant difference in C_p (t = 0.481, p = 0.632). Furthermore, at the specified threshold, the cerebellar functional brain networks of all participants displayed significantly higher clustering coefficients (gamma > 1) and nearly identical characteristic path lengths (lambda ≈ 1) relative to comparable random networks. These findings confirmed the presence of a typical small-world topology (sigma > 1) of the cerebellar brain network in both young and older adults. Nevertheless, in contrast to older adults, young adults showed significantly higher values of sigma (t = 3.450, p < 0.001) and gamma (t = 3.774, p < 0.001) (Fig. 4).

Fig. 4.

Differences in cerebellar small-world metrics between young and old adults. A. Violin plots were presented to show the AUC parameters of the clustering coefficient (C_p), characteristic path length (L_p), small-worldness (σ, sigma), normalized clustering coefficient (γ, gamma), and normalized characteristic path length (λ, lambda) for young and old adults, with means and standard deviations illustrated. *: p < 0.05. **: p < 0.01. ***: p < 0.001. B. C_p, L_p, sigma, gamma, and lambda values were shown across a density range from 10 to 34%. *: p < 0.05

Differences in cerebellar regional nodal features between young and older adults

When comparing young and older adults, significant differences were observed in regional nodal features. Young adults showed an increase in BC within the visual network (VN) compared to older adults (p = 0.042, FDR-corrected). In terms of NE, young adults showed significant increases in the VN and lateral somatomotor network (SMN(Lat)) compared to older adults (p = 0.038 for both, FDR-corrected). Additionally, young adults exhibited higher NLE than older adults in the DMN, and SMN(Lat) (p = 0.013 and p = 0.030, respectively, FDR-corrected). These findings were detailed in Fig. 5 and Table 1.

Fig. 5.

Differences in cerebellar regional nodal features between young and old adults. A. Degree centrality (DC). B. Betweenness centrality (BC). C. Nodal efficiency (NE). D. Nodal local efficiency (NLE). Green spheres represent nonsignificant nodes. Red and blue spheres indicate significant group differences (red: young > old; blue: young < old; FDR-corrected). The size of significant nodes corresponds to the effect size of the group differences

Table 1.

Alterations in regional nodal features between young and old adults

Brain Regions	ROI	Anatomical Classification	Peak MNI coordinate			p-value
			X	Y	Z
Betweenness centrality
VN	10	Vermis VⅠ	0	−74	−25	0.042
Nodal efficiency
VN	10	Vermis VⅠ	0	−74	−25	0.038
SMN(Lat)	25	Right Ⅴ	10	−62	−18	0.038
Nodal local efficiency
DMN	6	Left Cruse Ⅰ	−24	−76	−28	0.013
SMN(Lat)	25	Right Ⅴ	10	−62	−18	0.030

The p-values were obtained from two-sample t-tests with false discovery rate (FDR) correction (p < 0.05), evaluated across a range of network thresholds (0.10–0.34, in 1% increments). The numbers in the second column correspond to the twenty-seven cerebellar seed points based on the Seitzman atlas. The third column indicates the anatomical classification derived from the Spatially Unbiased Infratentorial Template (SUIT). Abbreviations: DMN, default mode network; MNI, Montreal Neurological Institute; ROI, region of interest; SMN(Lat), somatomotor network-lateral; VN, visual network

Correlations between cerebellar network metrics and cognitive performance

Partial correlation analysis was conducted to assess the relationships between cerebellar network properties and performance on the Stroop task in older adults. The gamma value of the cerebellar network was positively correlated with the ACC in the incongruent condition (r = 0.370, p = 0.035) in older adults (Fig. 6). However, no significant correlations were found between other global and regional topological metrics and performance on the Stroop task.

Fig. 6.

Correlation between cerebellar global metrics and Stroop test accuracy (incongruent) in older adults. Abbreviations: ACC, accuracy

Differences in whole-brain global properties between young and older adults

For network efficiency, young adults demonstrated significantly higher values of E_glob (t = 3.108, p = 0.003) and E_loc (t = 3.017, p = 0.003) compared to older adults (Fig. S2A and B). Regarding modularity analysis, there was no significant difference in the modularity index (Q) between young and older adults (t = 0.606, p = 0.546) (Fig. S3A and B).

As for small-world metrics, older adults exhibited a significantly higher value of L_p compared to young adults (t = −2.988, p = 0.004), while no significant difference was observed in C_p (t = 1.137, p = 0.259). Furthermore, at the specified threshold, the whole-brain networks of all participants confirmed the presence of a typical small-world topology (sigma > 1). However, statistical analyses indicated no significant differences between young and older adults in sigma (t = 1.471, p = 0.145), gamma (t = 1.477, p = 0.143), and lambda (t = 0.552, p = 0.582) (Fig. S4).

Differences in whole-brain regional nodal features between young and older adults

Across the whole brain, older adults exhibited significant alterations in DC at 48 nodes. Among them, 6 nodes were located in the cerebellar region, with 4 nodes in the FPN showing higher DC in older adults compared to young adults (p < 0.010, FDR-corrected), and 2 nodes in the dorsal somatomotor network (SMN(Dor)) exhibiting lower DC in older adults (p < 0.010, FDR-corrected) (Table S2, Fig. S5A). For NE, 126 of the 300 nodes showed significant age-related changes (p < 0.050, FDR-corrected), with older adults showing lower NE compared to young adults across all these nodes. Among these, 2 nodes in the cerebellar region belonged to the SMN(Dor) (Table S2, Fig. S5C). Additionally, young adults showed higher NLE at 129 nodes (p < 0.050, FDR-corrected), including 4 cerebellar nodes (2 in the FPN and 2 in the SMN(Dor)) (Table S2, Fig. S5D).

Effects of resting-state conditions on age-related differences in cerebellar network topological properties

No significant interactions were observed between age groups (young vs. old) and resting-state conditions (eyes open vs. eyes closed) for any cerebellar network topological properties (all p > 0.05) (Table S3 and Fig. S6).

Sex-specific age-related differences in cerebellar network properties

Given the limited sample size, this study conducted exploratory analyses to investigate sex-specific age-related differences in cerebellar network properties. In males, significant age-related changes were found in three metrics: E_glob (t = 2.093, p = 0.043), L_p (t = −2.083, p = 0.044), and lambda (t = 2.393, p = 0.021). Compared with young adults, the cerebellar networks in older males exhibited lower global efficiency and normalized characteristic path length, yet a longer characteristic path length. Other metrics did not reach statistical significance (Table S4 and Fig. S7). By contrast, older females demonstrated more pervasive age-related degradation, with significant differences detected in E_glob (t = 2.755, p = 0.009), E_loc (t = 2.331, p = 0.025), Q (t = 3.451, p = 0.001), L_p (t = −2.790, p = 0.008), sigma (t = 3.895, p < 0.001), and gamma (t = 3.664, p < 0.001). Older females showed reduced global efficiency, local efficiency, modularity, small-world properties (sigma, gamma), and longer path length in cerebellar networks compared to young females (Table S5 and Fig. S8). Partial correlation analysis revealed a significant negative correlation between the lambda value of the cerebellar network and Stroop test reaction time under congruent conditions in older adult males (r = −0.712, p = 0.048) (Fig. S9), and a significant positive correlation between the gamma value and Stroop test accuracy under incongruent conditions in older adult females (r = 0.431, p = 0.040) (Fig. S10).

Discussion

The current study provides valuable insights into the age-related differences in cerebellar functional network topologies and their associations with cognitive performance. In this study, notable differences in cerebellar network topologies and regional nodal features between young and older adults were identified. Older adults showed decreased network efficiency, modularity, and small-world properties, and increased path length in the cerebellar network. This indicates that the cerebellar networks of older individuals are less efficient in processing and integrating information, exhibit weaker segregation and integration of functional modules, and have a diminished capacity for effective information transfer and network integration. At the nodal level, compared with young adults, older adults exhibited lower BC in the VN, significantly decreased NE in both the VN and SMN(Lat), and reduced NLE in the DMN and SMN(Lat). This suggests that the neural communication and functional integration within critical cerebellar networks are less efficient in older adults. Partial correlation analysis further revealed significant correlations between cerebellar network metrics and cognitive performance in older adults, indicating that cerebellar network properties are closely related to cognitive functions in older adults. Notably, aging reduced whole-brain (including cerebellar) network efficiency while preserving modularity and small-worldness; at the nodal level, cerebellar regions exhibited specific reductions in NE and NLE, coupled with reorganization of DC. Furthermore, exploratory analyses showed that the topological organization of cerebellar functional networks during aging differed between genders, with males having limited declines and females displaying more extensive deterioration.

Age-related degradation of cerebellar functional network topology

The topological organization of the human brain generally exhibits high levels of local and global efficiency. High local efficiency, characterized by densely interconnected clusters of topologically adjacent regions, enhances the brain’s fault tolerance and supports efficient information processing within specialized functional areas [19, 45]. Previous studies have revealed that older adults exhibit reduced local efficiency in brain functional networks when compared to young adults [22, 23]. Our study found that the same outcome occured in the cerebellar network: the local efficiency of the cerebellar network was reduced in older adults. Global efficiency quantifies the brain’s capacity for parallel information transfer and integrated processing [19]. A high global efficiency suggests the existence of an integrated network that enables rapid information exchange among distributed elements [19]. Existing studies have demonstrated that older adults have lower global efficiency compared to young adults [23, 46]. Notably, our research findings suggested that, similar to the overall brain functional networks, the global efficiency of the cerebellar network in older adults was also significantly reduced when compared to that in young adults. These findings collectively demonstrate that age-related declines in network efficiency represent a pan-cerebral phenomenon, extending beyond well-documented cortical alterations to include cerebellar circuitry. Additionally, modularity is a fundamental organizational principle of brain networks, characterized by distinct functional communities with dense internal connections [47]. Modularity promotes specialized and segregated information processing within brain networks [19]. Previous research has revealed that modularity shows a linear decline throughout the lifespan [22]. Consistent with this finding, our comparison of cerebellar modularity between young and older adults further revealed a significant decline in modularity among the elderly. This suggests that the aging process may have a substantial impact on the modular structure of the cerebellum, potentially affecting its functional integrity and information-processing capabilities. Small-world networks are characterized by high local clustering and short characteristic path lengths, supporting both specialized and integrated information processing [37, 48]. Research has shown that the brain networks of healthy individuals exhibit an economical small-world topology [48]. Moreover, the small-worldness of resting-state functional brain networks declines with aging [49]. The current findings revealed significant age-related alterations in the small-world properties of cerebellar functional networks. Specifically, older adults exhibited longer L_p than young adults, indicating less efficient neural processing in cerebellar networks with normal aging. This aligns with previous research showing that aging is associated with declines in long-range connectivity, which may impair the cerebellum’s ability to rapidly coordinate information across distributed neural systems [50]. Moreover, when comparing the small-worldness (sigma) of cerebellar networks between the two groups, a significant decline was found in older adults. Small-worldness measures the equilibrium between neural integration and segregation [37]. In older adults, this delicate balance is disrupted, potentially affecting the overall efficiency and adaptability of cerebellar networks. Older adults also showed lower normalized clustering coefficient (gamma), indicating impaired segregation within neural networks. This is consistent with the overall decline in the small-world characteristics of the aging brain, reinforcing the notion that aging leads to functional alterations in neural networks.

In the current study, the cerebellar nodal features of the VN, DMN, and SMN(Lat) in older adults exhibited distinct age-related alterations. Specifically, older adults showed a significant reduction in BC within the cerebellar VN, especially in the Vermis Ⅵ region (as delineated by the Spatially Unbiased Infratentorial Template (SUIT)) [32, 51]. BC quantifies the extent to which a node functions as a crucial bridge for information transfer between different network modules [38]. Thus, the decreased BC indicates that the capacity of the Vermis Ⅵ region to promote seamless information flow has been weakened, potentially leading to reduced efficiency in motor coordination, working memory, and executive functions [52]. This decline is probably related to the reduction in cortical surface area and GM volume in the Vermis Ⅵ region during normal aging [53, 54]. Moreover, NE was significantly decreased in both the VN (Vermis VI) and the SMN(Lat) (Right Lobule V). These regions are critical for motor coordination, as well as cognitive functions such as working memory and executive function [52]. The reduced NE in these regions suggests a decline in information transmission efficiency at these function-related neural network nodes [38, 39]. This decline may be linked to age-related functional changes in motor coordination, cognitive processes, and emotional regulation. Furthermore, NLE was significantly reduced in both the DMN and SMN(Lat) in older adults [39]. The DMN, a brain network activated during rest, is intrinsically associated with self-reflection, mental exploration, and autobiographical memory [55]. The SMN(Lat) is responsible for fine sensorimotor control of the limbs [32]. These alterations correspond to Left Crus I, a region critical for working memory, executive function, and affective regulation, and Right Lobule V, which supports motor and somatosensory functions [52]. Reduced NLE suggests compromised local network resilience, which may underlie age-related deficits in cognitive-emotional processing and sensorimotor integration. Collectively, these findings reveal changes in BC, NE, and NLE across the VN, SMN(Lat), and DMN, highlighting their functional importance in aging. Specifically, degradation of the VN may impair motor coordination and cognitive function, alterations in the SMN(Lat) may reduce motor precision and sensorimotor coordination, and DMN aging may contribute to declines in higher-order cognitive functions. These results suggest that the aging of the cerebellar network is characterized by hub function degradation, global information transmission efficiency decline, and local network resilience impairment, presenting a multi-level decline in efficiency [56].

Associations between cerebellar network metrics and cognitive performance in older adults

This study explored the relationship between cerebellar network properties and Stroop task performance in older adults through partial correlation analysis. The results revealed a unique association between cerebellar network gamma values and the accuracy of the Stroop task in the elderly under incongruent conditions, which provides a new perspective for understanding the neurocognitive interaction in the aging process. Under incongruent conditions, the Stroop task demands that individuals suppress dominant responses (e.g., automatic semantic reactions to color-related words) and accurately report the ink color. This process heavily relies on cognitive control mechanisms and the ability to resolve conflicts [57]. In this study, the significant positive correlation between cerebellar network gamma values and task accuracy under these conditions indicates that older adults with higher gamma values perform better in handling cognitive conflicts. This finding aligns with previous research emphasizing the role of the cerebellum in cognitive control: the cerebellum not only participates in motor regulation but also plays a crucial role in executive functions through extensive connections with cortical regions such as the frontal and parietal lobes [43, 58, 59].

Age-related vulnerability of the cerebellum in whole-brain functional network topology

In terms of whole-brain network efficiency (including the cerebellum), our findings demonstrated a significant reduction in both global and local efficiency in older adults. This suggests age-related declines in the brain’s capacity for whole-network information transfer and local-level information exchange [23]. These results align with previous research [22, 23, 46]. Regarding modularity, no significant difference was observed between young and older adults at the whole-brain level, contrasting with findings of the cerebellar region. This aligns with the findings of Meunier et al., 2010 indicating that the global brain maintains modularity through functional reorganization during aging, whereas the cerebellar network organization exhibits greater age-related vulnerability [47]. In terms of small-world metrics, older adults exhibited longer L_p in the whole brain, reflecting less efficient global integration, while the clustering coefficient remained unchanged, paralleling observations in the cerebellum. Moreover, small-world parameters (sigma, gamma, and lambda) showed no significant group differences in the whole-brain network, whereas significant declines in sigma and gamma were observed in the cerebellar network. This finding is consistent with the results of Panhwar et al., 2022 on age-related differences in structural network properties, demonstrating that aging has no significant impact on the small-world attributes of the structural brain network [60]. This dissociation suggests that the cerebellum is more vulnerable to age-related alterations in network segregation and integration.

At the whole-brain nodal level, 48 nodes exhibited significant age-related alterations in DC. Notably, DC increased in cerebellar FPN nodes, possibly indicating compensatory hub recruitment in response to cognitive decline [38, 39]. This finding further supports extending the scaffolding theory of aging and cognition (STAC) to include the cerebellum, suggesting that the widespread increase in frontal activation with age may reflect an adaptive brain mechanism engaging compensatory scaffolding to counteract the challenges posed by neural decline [61, 62]. In contrast, DC decreased in the dorsal somatomotor regions, consistent with known age-related declines in motor and sensory integration [63]. Furthermore, older adults exhibited reductions in NE, reflecting impaired parallel information transfer, in cerebellar regions of the SMN(Dor). Additionally, NLE, indicating diminished local network resilience, was found in four cerebellar nodes located within both the FPN and SMN(Dor), which may be associated with cognitive and motor impairments in aging [64]. In summary, these findings reveal that despite compensatory reorganization in some cerebellar regions within the whole-brain network, the aging cerebellum is primarily characterized by reduced information processing efficiency and impaired local network integrity, underscoring its unique vulnerability in aging.

Sex-specific patterns of age-related cerebellar functional network degradation

Through exploratory analyses, this study revealed sex-specific differences in cerebellar network properties during aging. Despite the limited sample size, the findings suggest that males and females may experience distinct patterns of age-related cerebellar functional decline. Male participants exhibited declines limited to E_glob and lambda, and increases in L_p. In contrast, females exhibited more extensive network functional deterioration, involving global and local network efficiency, modularity index, and small-world properties. Previous studies indicate that males exhibit larger absolute cerebellar GM volume than females across most lobules, which may protect against age-related degeneration [65, 66]. This finding may also help explain females’ higher susceptibility to dementia.

Strengths and limitations

The study is the first to comprehensively characterize age-related changes in the topology of the cerebellar functional network and regional nodal features. This fills a gap in our understanding of how the cerebellar functional architecture degenerates with age. Moreover, it further explores the relationship between these changes and cognitive function, providing valuable insights into the mechanisms underlying age-related cognitive decline. Additionally, the study performs exploratory analyses to examine sex-specific differences in cerebellar network properties during the aging process, offering insights into the differential effects of aging on males and females. However, several limitations should be acknowledged. Firstly, the cross-sectional nature of our study limits our ability to track intraindividual changes over time. Consequently, it could not fully capture the changes in cerebellar network characteristics across the lifespan. Secondly, while partial correlation analysis was employed to explore the relationship between cerebellar network properties and cognitive performance, it is inherently limited in establishing causality. Moreover, the evaluation of cognitive function utilized the MMSE and the Stroop task, which cannot encompass all cognitive domains. Furthermore, considering the limited sample sizes of male and female participants in both young and old age groups, the findings about gender-specific differences in cerebellar network properties during aging require cautious interpretation.

Conclusion

Our findings revealed significant disparities in topological structure and node features between young and older adults, suggesting that aging may be associated with a decline in the cerebellar network’s capacity for efficient information processing and integration. Additionally, the integrity of the cerebellar network played a crucial role in supporting cognitive control functions as individuals age. Exploratory analysis also revealed gender differences in cerebellar network topology during aging, with males experiencing limited declines and females showing more extensive deterioration. These findings emphasize that studying the characteristics of the cerebellar network is of vital importance for comprehending the mechanisms of cognitive aging and provide guidance for exploring subsequent intervention strategies.

Supplementary Information

Below is the link to the electronic supplementary material.

Authors’ contributions

Mingzhu Ye: Conceptualization, methodology, formal analysis, investigation, original manuscript writing, and visualization. Haishuo Xia: Investigation, data acquisition, and manuscript review & editing. Tao Song, Zijin Liu: Methodology and manuscript review & editing. Antao Chen: Conceptualization, supervision, resource provision, and manuscript review & editing.

Funding

This work was supported by the National Natural Science Foundation of China (Grant Nos. 32171040, 32371105).

Data availability

Given that further research is currently in progress, data sharing is not applicable for this article.

Declarations

Ethics and consent to participate

This study was approved by the Institutional Review Board at the Brain Imaging Center of Southwest University and adhered to the ethical guidelines of the Declaration of Helsinki. All participants provided written informed consent.

Competing interests

The authors declare no competing interests.