Discriminative analysis of resting-state brain functional connectivity patterns of Attention-Deficit Hyperactivity Disorder using Kernel Principal Component Analysis

Resting state fMRI is an emerging research area that can reveal disorders and dysfunction of human brain. Resting state functional connectivity patterns have been reported for diagnosis of different diseases. In this study, Kernel Principal Component Analysis (KPCA) method based on connectivity matrix of each functional meaningful brain regions was applied to discriminate adult with Attention-Deficit Hyperactivity Disorder (ADHD) from normal controls. Firstly, functional connectivity matrix was obtained as classify patterns. Secondly, Kendall tau rank correlation coefficient was applied to select features with high discriminative power, while KPCA was applied to find the abnormal pattern of ADHD in a much lower mainfold. Finally, the two groups of ADHD and normal participants were classified by a Support Vector Machine (SVM) classifier. Experimental results showed that SVM based on KPCA can produce a correct classification rate of 81% using a leave-one-out cross validation, which indicate that KPCA is an effective method that can greatly improve the final discriminative performance. Moreover, we examined the brain regions with statistically significant difference, like the frontal cortex, insula, cingulate cortex, postcentral gyrus, thalamus, middle temporal cortex, well confirmed with previous findings on ADHD. From the classification performance, we conclude that KPCA based on functional connectivity matrix can provide useful information for diagnosis of ADHD and even other diseases.