Computer Aided Diagnosis of ADHD Using Brain Magnetic Resonance Images

摘要

This paper presents a pilot study on the development of an automated diagnostic tool for Attention Deficiency Hyperactivity Disorder (ADHD) based on regional anatomy of the child brain. For the pilot study, amygdala and cerebellar vermis are chosen from magnetic resonance images obtained from ADHD-200 consortium data set. These regions play a vital role in the control of emotional response and behavior/locomotion, respectively. The images are preprocessed, registered by transforming each image to the space of the population average. The gray matter tissue probability values of amygdala and cerebellar vermis are obtained by applying a region-of-interest mask. These values are then used to train a Projection Based Learning algorithm for a Meta-cognitive Radial Basis Function Network (PBL-McRBFN) for the diagnosis of ADHD and prediction of its subtype. Performance results show that the PBL-McRBFN diagnoses ADHD and predicts its subtypes based on these regions with an accuracy of approx. 65% and 62%, respectively.