Diagnosis of attention deficit hyperactivity disorder using deep belief network based on greedy approach

Abstract

Attention deficit hyperactivity disorder creates conditions for the child as s/he cannot sit calm and still, control his/her behavior and focus his/her attention on a particular issue. Five out of every hundred children are affected by the disease. Boys are three times more than girls at risk for this complication. The disorder often begins before age seven, and parents may not realize their children problem until they get older. Children with hyperactivity and attention deficit are at high risk of conduct disorder, antisocial personality, and drug abuse. Most children suffering from the disease will develop a feeling of depression, anxiety and lack of self-confidence. Given the importance of diagnosis the disease, Deep Belief Networks (DBNs) were used as a deep learning model to predict the disease. In this system, in addition to FMRI images features, sophisticated features such as age and IQ as well as functional characteristics, etc. were used. The proposed method was evaluated by two standard data sets of ADHD-200 Global Competitions, including NeuroImage and NYU data sets, and compared with state-of-the-art algorithms. The results showed the superiority of the proposed method rather than other systems. The prediction accuracy has improved respectively as +12.04 and +27.81 over NeuroImage and NYU datasets compared to the best proposed method in the ADHD-200 Global competition.