Table 1.
Demographic summaries for all datasets provided by the ABIDE Consortium.
SD: standard deviation.
Fig 1.
First, functional connectivity (FC) networks were constructed by estimating the Pearson’s correlation coefficient between ROIs. Second, the spectral clustering algorithm was applied to the average functional network to obtain the sub-networks (clusters). Then, the network fitness index (NFI) was calculated for each ROI. To identify the ROIs associated with age, we modeled the NFI using a generalized linear model with the number of volumes removed by the “scrubbing” procedure as a covariate. This procedure was applied to data with and without “scrubbing” per site. P-values of different sites were combined by using the Fisher’s method. ROIs that presented a difference in nominal p-values greater than 5% were excluded. The remaining p-values were corrected for false discovery rate. ROIs with a corrected p-value lower than 5% in the scrubbed data were considered to be associated with age.
Fig 2.
Application of the silhouette statistic to estimate the number of sub-networks.
The x and y axes represent the number of clusters and the silhouette statistic, respectively. Note that the largest silhouette statistic is obtained when the number of clusters is equal to five. In other words, the silhouette statistic suggests clustering the brain functional network into five sub-networks.
Fig 3.
Three-dimensional views and sagittal slices of the entire brain.
The five sub-networks obtained by the spectral clustering algorithm, namely, cerebellar, default mode, fronto-parietal, somatomotor, and visual, are discriminated by different colors. L: Left; R: Right.
Fig 4.
The left middle temporal gyrus and the left putamen are highlighted in different colors. L: left; R: right.
Fig 5.
Fitted line obtained for the left middle temporal gyrus and the left putamen.
The x and y axes represent the age in years and the network fitness index (NFI), respectively. Note that the NFIs are negatively associated with age in both ROIs. The slopes are −0.005 with p < 0.001 for the left middle temporal gyrus and −0.007 with p < 0.001 for the left putamen.
Fig 6.
The proportion of subjects that present the left middle temporal gyrus and the left putamen assigned to a specific cluster in each neurodevelopmental stage (children, adolescents, and adults).
CeN: cerebellar network; DMN: default mode network; FPN: fronto-parietal network; SN: somatomotor network; and VN: visual network.
Fig 7.
The average and 95% confidence intervals (represented by error bars) of the network fitness index for the left middle temporal gyrus and the left putamen in each developmental stage.
The error bars represent the 95% confidence intervals.