Detection of atypical network development patterns in children with autism spectrum disorder using magnetoencephalography

Abstract

Autism spectrum disorder (ASD) is a developmental disorder that involves developmental delays. It has been hypothesized that aberrant neural connectivity in ASD may cause atypical brain network development. Brain graphs not only describe the differences in brain networks between clinical and control groups, but also provide information about network development within each group. In the present study, graph indices of brain networks were estimated in children with ASD and in typically developing (TD) children using magnetoencephalography performed while the children viewed a cartoon video. We examined brain graphs from a developmental point of view, and compared the networks between children with ASD and TD children. Network development patterns (NDPs) were assessed by examining the association between the graph indices and the raw scores on the achievement scale or the age of the children. The ASD and TD groups exhibited different NDPs at both network and nodal levels. In the left frontal areas, the nodal degree and efficiency of the ASD group were negatively correlated with the achievement scores. Reduced network connections were observed in the temporal and posterior areas of TD children. These results suggested that the atypical network developmental trajectory in children with ASD is associated with the development score rather than age.

Fig 1. Association matrix and difference between groups.

(A) The average association matrix in children with ASD. (B) Subtraction of the average association matrix in TD children from that in the children with ASD. (C) The connections showing significantly different (p < 0.05) mutual information between the ASD and TD groups.

Fig 2. Mean brain networks of the two groups.

The mean brain networks of (A) children with ASD and (B) TD children were generated by binarizing the mean association matrices to the adjacency matrices with an equal number of 1. The degree of (A) and (B) was 23. (C) The dashed lines indicate specific connections that exist only in the mean brain networks of children with ASD, and the solid lines indicate connections that exist only in the mean brain networks of TD children.

Fig 3. Network level indices as a function of degree.

(A) The characteristic path length (L) and (B) the clustering coefficient (C) in children with ASD (asterisks) and TD children (triangles) as a function of the average degree of the network. The error bars correspond to standard deviations. The corresponding values of the regular and random networks are indicated by dashed and solid lines, respectively.

Fig 4. Correlation coefficients between network level indices and the participants’ age.

Panels (A) and (B) show the correlation coefficients between L and C, respectively, as well as the age of the children in the ASD group as a function of the degree. The correlation coefficients between L and C and the age of the children in the TD group are shown in (C) and (D), respectively. The shaded bars show the intervals at which the graph indices and age were significantly correlated (p < 0.05).

Fig 5. Correlation coefficients between network level indices and the achievement scores.

The L and C results of ASD group are shown in panels (A) and (B). Panels (C) and (D) correspond to the results from the TD group. The significantly correlated intervals (p < 0.05) are also indicated by shaded bars.

Fig 6. Comparison between nodal level indices in children with ASD and TD children.

The network degree was set at 23. Panels (A) and (B) show the maps of the values of the z-statistic. Warm colors indicate high median values in the TD group, and the cool colors represent high median values in the ASD group. Panel (C) shows the p-value map showing significant differences (p < 0.05). Blue dots indicate the regions in which the p-values are between 0.01–0.05.

Fig 7. Correlation maps between the nodal level indices and the achievement scale.

The network degree was set at 23. Panels (A-D) show the maps of the correlation coefficients between the nodal level indices and the age of the children in the ASD (A, B) and TD (C, D) groups. Panels (A) and (C) show the correlation coefficients between the nodal degree and the achievement scale. Panels (C) and (D) show the correlation coefficients between the nodal efficiency and the achievement scale. The corresponding p-value maps (E-H) are shown under each correlation map. Red dots indicate the regions in which the p-values of the nodes are less than 0.01. Blue dots indicate the regions in which the p-values are between 0.01–0.05.

Fig 8. The correlation coefficients between the nodal level indices for two typical sensors and the raw scores of the achievement scale as a function of degree.

Panels (A) and (B) show the NDPs of the ASD group on the sensor 131. The NDPs of the TD group at sensor 10 are shown in panels (C) and (D). A strong negative correlation can be observed in both groups. The shaded bars indicate p-values less than 0.05. The sensor locations are shown in panel (E).