Limbic/paralimbic connection weakening in preschool autism-spectrum disorder based on diffusion basis spectrum imaging.

Abstract

This study aims to investigate the value of basal ganglia and limbic/paralimbic networks alteration in identifying preschool children with ASD and normal controls using diffusion basis spectrum imaging (DBSI). DBSI data from 31 patients with ASD and 30 NC were collected in Hunan Children's Hospital. All data were imported into the post-processing server. The most discriminative features were extracted from the connection, global and nodal metrics separately using the two-sample t-test. To effectively integrate the multimodal information, we employed the multi-kernel learning support vector machine (MKL-SVM). In ASD group, the value of global efficiency, local efficiency, clustering coefficient and synchronization were lower than NC group, while modularity score, hierarchy, normalized clustering coefficient, normalized characteristic path length, small-world, characteristic path length and assortativity were higher. Significant weaker connections are mainly distributed in the limbic/paralimbic networks. The model combining consensus connection, global and nodal graph metrics features can achieve the best performance in identifying ASD patients, with an accuracy of 96.72%.The most specific brain regions connection weakening associated with preschool ASD are predominantly located in limbic/paralimbic networks, suggesting their involvement in abnormal brain development processes. The effective combination of connection, global and nodal metrics information by MKL-SVM can effectively distinguish patients with ASD.