Diagnosis of Autism Spectrum Disorder Using Age-Specific Brain Morphological Features

Autism Spectrum Disorder (ASD) diagnosis based on sMRI can be more objective than clinical scales due to high heterogeneity. However, accuracies on large heterogeneous datasets were not high. We used age-specific features based on ABIDE II dataset to distinguish ASD and control. In the meanwhile, we combined two kinds of age-specific structural features including regional and interregional features using multiple kernel learning (MKL) to complement each other. Results showed that our procedure achieved accuracy of more than 85 percent on discriminating ASD from control.