Brain Function Network Analysis of Children with Attention-Deficit/Hyperactivity Disorder Based on Adaptive Sparse Representation Method

Abstract

Attention-Deficit Hyperactivity Disorder (ADHD), as a neuro-developmental disorder, has a great impact on children's life. Brain function network analysis is one of the crucial means for ADHD diagnosis. To better understand the ADHD, the method combined with Adaptive Sparse Representation (ASR) method and graph theory was proposed to achieve the global functional network. First, ASR was applied to calculate the correlation to construct the brain function network. Second, the obtained optimal threshold based on an absolute selection strategy aimed to reduce the weak correlative connections. In order to understand the differences between the ADHD and the normal, graph theory was utilized for brain network evaluation. The connection sensitivity of ASR of the simulated data is 88.89% shows the validity of the proposed method. The experiments were conducted on clinical resting-state fMRI data of 30 ADHD patients and 30 normal persons. Compared to the normal, the average shortest path of the ADHD was 24% higher, the average degree distribution of the ADHD was 21% lower, the local efficiency of the ADHD was 14% higher and the global efficiency of the ADHD was 19% lower. Meanwhile, there were significant differences of the node efficiency between the ADHD and the normal in the temporal lobe and occipital cortex. The experimental results showed that the proposed method could show more clearly the differences between the normal and the ADHD.