Differences in Brain Cortical Morphological Networks Between Gyri and Sulci

Introduction

The human brain, as an interconnected complex network, can be mapped through virtual imaging using multimodal magnetic resonance imaging (MRI) technology. By analyzing this network using graph theory methods, many studies have discovered some non-trivial topological features of the brain network, such as small-world organization, modular structure, and highly connected hubs. These findings provide important insights into the organizational principles of brain connectivity. However, these topological features are not uniformly distributed across the cerebral cortex. For example, significant differences in small-world organization exist between the left and right hemispheres. Moreover, a more important factor, the highly convoluted cortical folding pattern, has also been found to influence brain network topology.

The cortical folding pattern, consisting of protruding gyri and depressed sulci, is one of the most prominent features of the human brain structure. Evidence from structural and functional MRI studies suggests that brain networks formed by gyri have stronger connectivity and higher network efficiency compared to those formed by sulci. Despite these advances, research on the impact of gyri and sulci on brain structural networks derived from structural MRI data remains limited.

In this study, we aim to comprehensively analyze the influence of gyri and sulci on individual subject morphological brain networks. Specifically, we adopted previously developed methods to construct four types of individual subject morphological brain networks and analyzed them using multiple independent datasets. We examined the differences between gyral and sulcal morphological brain networks from multiple aspects, including inter-regional morphological similarity, small-world organization, test-retest reliability, explanatory power for behavior and cognition, and sensitivity to major depressive disorder (MDD).

Research Source

This study was jointly conducted by Qingchun Lin, Suhui Jin, Guole Yin, Junle Li, and other collaborators from the Brain Research and Rehabilitation Institute of South China Normal University and various universities and research institutions. The paper was published in the journal “Neuroscience Bulletin” and accepted on March 28, 2024.

Research Methods

Participants and Data Collection

The study included four independent datasets:

1. Human Connectome Project (HCP) dataset: Included 1113 participants who underwent T1-weighted structural MRI scans. 444 participants were included, with an average age between 22-35 years.
2. Beijing Normal University (BNU) TRT dataset: Included 57 healthy participants who underwent two MRI scans, with an average interval of 40.94 days.
3. Southwestern University (SWU) TRT dataset: Included 121 healthy participants who completed three MRI scans, with intervals ranging from 120 to 653 days.
4. MDD dataset: Included 100 first-episode, drug-naive patients with major depressive disorder and 99 healthy controls.

Image Preprocessing

The CAT12 toolbox was used for surface-based vertex analysis of individual structural images, extracting four morphological features: cortical thickness (CT), fractal dimension (FD), gyrification index (GI), and sulcus depth (SD). Following the toolbox manual, 2D CT images were smoothed, while other images were smoothed using a Gaussian kernel based on their nature.

Construction of Individual Sample Morphological Brain Networks

The Destrieux atlas was used to divide the cortical surface into 74 regions of interest (ROIs) per hemisphere, ultimately retaining 60 regions per hemisphere after excluding ambiguous areas. Then, by calculating the distribution of morphological features for each ROI, inter-regional morphological similarity was estimated using the Jensen-Shannon divergence (JSD) method, resulting in four morphological brain networks: CTN, FDN, GIN, and SDN.

Network Analysis

In the network analysis, a proportional threshold method was applied to the HCP dataset to exclude low similarity edges. Small-world parameters (including clustering coefficient Cp and characteristic path length Lp) were calculated and normalized through generated random networks. Additionally, intraclass correlation coefficient (ICC) was used to evaluate short-term and long-term test-retest reliability. For behavioral and cognitive association analysis, a multivariate variance component model was used to assess the explanatory power of morphological brain networks for cognitive and behavioral differences, followed by edge-wise correlation analysis.

Research Results

Morphological Similarity Differences Between Gyri and Sulci

CTN and GIN showed significantly higher morphological similarity in gyrus-gyrus networks compared to sulcus-sulcus networks, while FDN showed higher similarity in sulcus-sulcus networks. Significant differences in morphological similarity between gyri and sulci were observed within various regions, especially for CTN and GIN, showing higher morphological similarity among gyri.

Differences in Small-World Parameters

Both gyral and sulcal networks exhibited small-world organization, but gyral networks showed significantly lower clustering coefficients and higher characteristic path lengths in CTN and GIN compared to sulcal networks. These results suggest that gyral networks are more prominent in functional integration.

Differences in Test-Retest Reliability

Both gyral and sulcal networks showed high test-retest reliability regardless of time intervals. However, over longer time intervals, sulcal networks displayed significantly higher reliability in CTN, GIN, and SDN, while showing the opposite result in FDN.

Ability to Explain Behavioral and Cognitive Differences

Morphological brain networks significantly explained individual differences in cognitive and motor domains. In particular, gyral and sulcal networks and connections made important contributions to these associations, indicating the importance of gyral regions in supporting cognitive and motor functions.

Sensitivity to Major Depressive Disorder (MDD)

Only sulcus-sulcus networks showed significant decreases in morphological similarity in CTN, FDN, and GIN in MDD patients, suggesting that sulcal networks are more susceptible to MDD influence.

Conclusion

This study systematically explored the influence of gyri and sulci on individual subject morphological brain networks. The results showed significant differences between gyral and sulcal networks in multiple aspects such as morphological similarity, small-world organization, test-retest reliability, behavioral and cognitive associations, and sensitivity to MDD. These findings deepen our understanding of the impact of cortical folding patterns on brain network organization.