Subjects
Sixteen overweight or obese children (11.1±1.9 years; BMI 29.5±5.5 kg/m2) and twenty-one controls (11.5±1.9 years; BMI 18.14±2.3 kg/m2) volunteered for this study with parental consent (Table 1). All the recruited children were born clinically healthy and right-handed. The overweight/obese group (OC) was defined by a BMI equal to or greater than 24 kg/m2, while the BMI of the control group (HC) was lower than 24 kg/m2. Participants were excluded if they presented with any neurological abnormalities, mental deficiencies, psychological disorders, history of cerebral hemorrhaging or brain surgery, consumption of psychoactive drugs, and/or visible brain lesions.
Data acquisition and image processing
All subjects underwent brain scanning using a 3.0 T MR scanner to acquire 3D T1-weighted (T1W) images. Anatomical brain segmentation was performed on the T1W images using AccuBrain, an automated brain segmentation and volumetric quantification software [8]. Volumetric analysis was performed for brain parenchyma and anatomical brain structures such as subcortical (bilateral hippocampus, amygdala, thalamus, caudate, putamen, pallidum, ventral diencephalon (DC) and accumbens) and brainstem structures (midbrain, superior cerebellar peduncle (SCP), pons, medulla). The volumes of these structures were normalized by intracranial volume (ICV) and the resulting volume ratios (% of ICV) were used in the following statistical analyses.
Statistical analysis
The volumetric measures were compared between OC and HC groups using the non-parametric Mann-U Whitney Test. False discovery rate (FDR) correction was performed when needed regarding the multiple comparisons involved among the different brain volumetric measures.
To map the network-level of brain structural differences between OC and HC groups, we performed volumetric structural covariance analyses[9]. Here, the brain volumetric structural covariance (or anatomic connection) was defined as the correlation between the regional brain volumetric measures of different brain regions, where the correlation was measured using Spearman's rank partial correlation analyses. In detail, the correlations, which served as edges connecting the nodes of regions of interest (ROIs), were calculated between brain regional volumes of the selected regions in the OC group and in the HC group separately (with age and gender as the covariates). Only the significant correlation between two ROIs would be displayed with an edge between the nodes, and the edge was encoded with colors with respect to the magnitude of the corresponding correlation.
Specifically, we defined an ROI-ROI correlation as an additional connection if it was significant (after FDR correction) in the OC group but not in the HC group. Likewise, if an ROI-ROI correlation was significant in the HC group but not in the OC group, it was defined as a missing connection in the OC group. The additional connections in the OC group indicated the synergy volume alteration between the regions, and the missing connections in the OC group denoted the disconnection of these regions. In this study, we focused on the subcortical and brainstem structures as the ROIs for the structural covariance analyses.