Obesity is one of the most serious public health concerns among adults and children in the 21^st century, which increases risk of many other diseases, including cardiovascular risks, hypertension, dyslipidemia, endothelial dysfunction, etc., and it is commonly measured with body mass index (BMI) [1]. Previously studies about obesity mainly used diffusion tensor imaging (DTI) to examine the relationship between BMI and DTI parameters, and found that white matter integrity was reduced in obesity [2, 3]. However, the research about the particular structural brain network change of obese patients was tended to be less. Hence, our study aimed to map the structural connectomic changes over obese patients based on DTI tractography using graph theoretical and network-based statistic (NBS) analyses. The white matter tracts can be reconstructed by DTI tractography, and structural network of the entire brain can be obtained using graph theoretical analysis. Graph theoretical analysis can also be used to quantify differences between patient groups and appropriate comparison groups, and theoretically described the disruption or abnormal integration of spatially distributed brain regions [4]. The changes of white matter tracts that link regions throughout the brain can be calculated with network-based statistic analysis.

Brain DTI images from 50 participants, including 20 obese patients (BMI = 37.9 ± 5.2) and 30 healthy controls (HC) (BMI = 22.6 ± 3.4), were obtained with 1.5T MRI (Ingenia, Phillips, Netherlands). DTI parameters were TR/TE = 3279/110 ms; resolution = 3 x 3 mm²; slice thickness = 3 mm; diffusion orientations = 67; b-values = 0, 1000, 2000 s/mm², and 45 axial contiguous slices. The raw diffusion data for each participant were first corrected eddy current distortions using FMRIB (functional magnetic resonance imaging of the brains) Software Library (FSL). Each participant’s diffusion images were spatially normalized to the Montreal Neurological Institute (MNI) T2W template using parameters determined from the normalization of the diffusion null image to the T2W template using Statistical Parametric Mapping (SPM). Images were resampled with a final voxel size of 2 x 2 x 2 mm³. DSI Studio was performed for whole-brain DTI tractography, and the individual structural connectivity matrix of each participant with size of 90 x 90 could be establish establish by importation of ROIs based on the Automated Anatomical labeling (AAL).

Graph theoretical analysis was applied to investigate structural alteration of whole brain network, and the changes of regional and global topological organization in obese patients. The topological properties included clustering coefficient (C), normalized clustering coefficient (γ), local efficiency (E_local), characteristic path length (L), normalized characteristic path length (λ), global efficiency (E_global), small worldness index (σ), transitivity, assortavity, and modularity. The NBS was finally performed to find the significant sub-networks differences between the obese patients and HCs based on two-sample t-test.

Our results showed that small world topology was observed in both obese patients and HCs. We found significant lower E_global, E_local and transitivity in the obese patients compared with HCs (p<0, 05) (Fig. 1), while no significant difference between two groups was found in other topological measurements. The results indicated that the obese patients with a poor ability of local segregation and global integration, and also with a less transitivity in whole brain network compared to the HCs. In the visualization of structural network, the edge connections decreased in the obese patients (Fig. 2A) compared with HCs (Fig. 2B). The obese patients were considered as an abnormal connectivity network, which may lead to related disease. According to NBS result, a disrupted sub-network consisted of 40 regions and 47 edges was identified (P<0.05) (Fig. 2C). The disrupted edges in the obese patients were mainly distributed over left to right regions of the brain, which was the corpus callosum region.

In summary, the major finding in our study was the altered whole-brain structural topological organization in the obese patients, including poor local segregation, poor global integration and less transitivity. The second major finding in our study was altered structural connectivity in the obese patients mainly regarding corpus callosum region, which was considered as core region associated with obese patients. The decreased structural connectivity among the regions causing possibly local dysfunction might reflect the underlying mechanism in obesity.

In the result of graph theoretical analysis, poor ability of local segregation, global integration, and transitivity in the obese patients was found. In the result of NBS, decreased connections in structural connectivity network, and alterations in the corpus callosum region was observed. It may facilitate the understanding of underlying mechanism in obesity.