Neural Network Classification of Resting State fMRI Data of ADHD and Healthy Subjects using Dynamic Mode Features

Attention deficit hyperactivity disorder (ADHD) is one across the board mental disorder distinguished by inattentiveness, hyperactivity or impulsive behavior. The exact cause is yet unknown but the prevalence of ADHD requires early diagnosis and treatment. This study investigates the decomposed subsystems of resting- state fMRI (rs-fMRI) blood oxygen level dependent (BOLD) signals in order to assess the overall characteristic of the human brain. rs-fMRI BOLD signals were first decomposed into dynamic modes (DMs) which can illuminate the patterns of brain subsystems. Each DM is associated with one Eigenvalue that characterizes functional connectivity dynamics. Thereby the features related to those DM were extracted. Using the features obtained, the classification is performed using neural network-based pattern recognition model and the performance is evaluated. The interpreted results with neural network (NN) classifier obtains an accuracy of 73.7%.

I. Introduction

The credit to development of fMRI goes to Seiji Ogawa and Ken Kwong, which is the latest in long line of innovations, including positron emission tomography (PET) and near infrared spectroscopy (NIRS), which uses blood flow and oxygen metabolism to infer brain activity in late 1990's. Resting state fMRI measures brain activity when a person is at rest and not performing a specific task. Resting-state functional Magnetic