Differentiation between resting-state fMRI data from ADHD and normal subjects: Based on functional connectivity and machine learning

Attention-deficit/hyperactivity disorder (ADHD) is a neuropsychiatric disorder which is quite common in childhood, with an estimated prevalence of 5–8%, and often persists into adolescence and adulthood. It is further characterized as inappropriate developmentally symptoms of inattention, impulsiveness, motor over-activity and restlessness. The aim of this study is to evaluate the feasibility of diagnosing ADHD by analyzing the resting-state functional magnetic resonance imaging (fMRI) data. In addition to confirming the previously observed three areas including anterior cingulate cortex (ACC), posterior cingulated cortex (PCC) and ventro medial prefrontal cortex (vmPFC), we also found significant differences in cerebellum, motor cortex and temporal lobe between ADHD and normal humans based on regional homogeneity analysis of the dataset from 73 children with ADHD and 76 normal children. Extracting features from these seven brain areas and utilizing the LDA classifier, the average accuracy of distinguishing normal and ADHD children reaches 80.08% though 50 times of 2-fold validation. Experimental results demonstrate the feasibility of ADHD diagnosis based on the combination of functional connectivity of resting-state fMRI and machine learning technique.