Self-Brain Networks in Toddlers with Autism Spectrum Disorder

Abstract As early as 18-24 months, self-processing has been reported abnormal in autism spectrum disorder (ASD), echoing a key period for the emergence of self-awareness. However, the neural mechanisms underlying this phenomenon remains unclear. The self-processing was supposed integrating three nested levels: the interoceptive-self which represents bodily internal state, the exteroceptive-self which links the internal and external environment and marks a key step in self-awareness development, the mental-self which represents the mental world built on the abstract relationship between the internal and external environment. Therefore, we assumed that ASD may show abnormal exteroceptive-self from 18-24 months and then develop the disorders of mental-self with their growing. To test this hypothesis, resting-state functional magnetic resonance imaging data from both ASD and typically developing controls (TDC) were collected and divided into two age groups: 18-24 months and 25-48 months. In ASD, functional connectivity revealed hyperconnectivity within exteroceptive-self network; graph theory analysis revealed atypical nodal topological properties in both exteroceptive- and mental-self network. Based on these findings, a support vector machine (SVM) model was trained to distinguish ASD from TDC for each age groups, and then validated with independent datasets. The SVM models achieved accuracy rates of 88.2% (training) and 75% (testing) for the 18-24 months group, and 86.7% (training) and 84.6% (testing) for the 25-48 months group while SVM models based on the non-self-related brain regions cannot achieve similar accuracy. These findings disclose the abnormal information processing within self-network in brains of toddlers with ASD, and provide potential neuroimaging markers for its early detection and future translation.