Graph Theory Analysis of Brain Network Topology Structure in Patients with Primary Angle-Closure Glaucoma

Research Background

Glaucoma is a global blinding eye disease characterized by optic nerve damage and elevated intraocular pressure (IOP) (Kang and Tanna 2021). Among various types of glaucoma, Primary Angle-Closure Glaucoma (PACG) is particularly prevalent in Asia (Chan et al. 2016). The etiology of PACG is complex, involving anterior chamber angle narrowing, crowding leading to closure (Sun et al. 2017), choroidal thickening (Zhou et al. 2013; 2014), and choroidal expansion (Kumar et al. 2008), ultimately leading to elevated IOP. Environmental factors such as lighting conditions can also affect pupil size, thereby impacting the anterior chamber angle. Additionally, the loss of retinal ganglion cells is a significant pathological feature of glaucoma, and studies have shown that damage to these cells can lead to optic nerve and tract atrophy, affecting the lateral geniculate body and optic radiations, ultimately causing degenerative changes in the visual cortex (You et al. 2021; Rossiter 2015; Chen et al. 2013).

Previous neuroimaging studies on glaucoma mainly focused on revealing structural and functional abnormalities in the brains of glaucoma patients, such as differences in total gray matter volume (Jiang et al. 2018), abnormal functional connectivity in local and distant regions (Fu et al. 2022), and frequency-dependent changes in neural activity (Jiang et al. 2019). Although these studies have made certain progress in revealing the neural mechanisms of glaucoma, research on brain functional networks in glaucoma patients remains relatively limited.

Modern neuroimaging technologies have significantly advanced research progress in glaucoma. Functional Magnetic Resonance Imaging (fMRI) has been widely used to explore neural mechanism changes involved in glaucoma. Previous studies using fMRI have observed extensive changes in brain structure and function in glaucoma patients, such as brain areas affecting the visual pathway (Jiang et al. 2019) and brain areas primarily affecting emotion (Chen et al. 2019).

Source Description

This paper was written by Ri-Bo Chen, Xiao-Tong Li, and Xin Huang, with Ri-Bo Chen and Xiao-Tong Li contributing equally. The research was conducted at the Department of Radiology of Jiangxi Provincial People’s Hospital, the First Affiliated Hospital of Nanchang Medical College, and the Queen Mary School of Jiangxi Medical College. This paper was published in 2024 by the “Brain Topography” journal under Springer.

Research Methods

This study used graph theory analysis techniques to construct functional brain networks from fMRI data, conducting global metrics, node metrics, modularity, and network statistical analysis on the brain networks of PACG patients.

Subjects and Participants

The study included 44 patients diagnosed with PACG and 44 healthy controls. These participants were from Jiangxi Provincial People’s Hospital, and the study had been approved by the ethics committee. Inclusion criteria for PACG patients included confirmed anterior chamber angle narrowing, presence of glaucoma-related visual field defects, no prior glaucoma treatment, no history of head injury, no neurological or psychiatric disorders, and no MRI contraindications. The healthy control group was matched for no known eye disease, no history of head injury, no neurological or psychiatric disorders, no MRI contraindications, and matched for age, gender, and education level with the PACG group.

Imaging Data Acquisition and Preprocessing

All participants were scanned in a resting state with closed eyes using a 3T MRI scanner equipped with an eight-channel head coil, using a gradient-echo echo-planar imaging sequence to capture the BOLD signal changes. Data processing was conducted using MATLAB 2013a, DPABI software, and the SPM12 toolbox, including steps for converting DICOM to NIFTI format, slice time correction, head motion correction, BOLD image registration, normalization to MNI space, Gaussian smoothing, linear detrending, and band-pass filtering.

Brain Network Construction

Using GRETNA software, the brain was divided into 90 regions based on the Automated Anatomical Labeling (AAL) template, and the Pearson correlation coefficient of the BOLD time series for each pair of nodes was calculated, generating a 90×90 correlation matrix for each participant, which was normalized using Fisher’s r-to-z transformation for subsequent group comparisons.

Network Analysis

Global and Node Metrics

Global and local metrics of the brain networks were examined over a sparsity range of 0.05 to 0.50, in intervals of 0.01, calculating the area under the curve (AUC) for network metrics at multiple sparsity levels to reveal changes in brain network topology. Global metrics (such as small-world properties, network efficiency) and node metrics (such as betweenness centrality, degree centrality, node efficiency) were calculated.

Modularity Analysis

The 90 brain regions were divided into 9 functional modules, quantifying connections within and between modules to reveal the connectivity properties of each brain region. Modularity analysis showed that only module 5 (primarily involving occipital regions) had significant differences in internal connections between PACG and healthy controls.

Connectivity Analysis

Network-based statistics (NBS) were used to quantify differences in functional connectivity strengths in brain networks, employing non-parametric permutation tests to improve the accuracy of statistical results.

Statistical Analysis

Clinical variables were compared using SPSS 16.0, and inter-group differences in global network parameters and regional node parameters were assessed with two-sample t-tests, including age, gender, education level, and head motion as covariates.

Verification Analysis

For reliability validation of the results, supplementary analysis using the AAL 116 template was conducted.

Research Results

Demographics and Visual Measurements

There were no significant differences in gender and age between the two groups, but Best Corrected Visual Acuity (BCVA) showed significant differences. PACG patients had lower visual abilities in both left and right eyes compared to healthy controls. See Table 1 for details.

Global Metrics Results

In brain network analysis, PACG patients and healthy controls showed no significant differences in small-world properties and network efficiency properties, indicating no significant changes in whole-brain network connectivity characteristics between the two groups.

Node Metrics Results

Node centrality and node efficiency in the left superior frontal gyrus, right superior frontal gyrus, and right middle occipital gyrus were significantly higher in PACG patients, while node centrality and node efficiency in the right superior temporal gyrus were significantly lower, indicating significant connectivity changes in these brain regions in PACG patients.

Modularity Analysis Results

In the analysis of connections within and between modules, module 5 (occipital region) in PACG patients showed a unique connection pattern, and the connection strength between module 1 and modules 7 and 8 showed significant differences, reflecting changes in functional connectivity patterns of specific brain regions.

NBS Analysis Results

In brain functional network connectivity, PACG patients exhibited significantly enhanced functional connectivity, mainly involving the prefrontal, occipital, and temporal lobes. However, functional connectivity strength in certain subcortical and temporal regions was significantly reduced.

Conclusion and Significance

Significant changes in node metrics and modularity were found in the functional brain networks of PACG patients, particularly in the frontal, occipital, temporal, and cerebellar regions. However, the overall connectivity pattern of the whole brain network did not show significant changes in PACG patients. The study results can serve as markers for early diagnosis and differentiation of PACG, and targeting interventions on brain regions with high centrality and node efficiency may help optimize treatment plans.

Future research could validate and explore brain network function changes in PACG patients with larger sample sizes and longitudinal studies, providing a more precise theoretical basis for clinical intervention and treatment.