International Conference on Computer and Knowledge Engineering

Home / 11th International Conference on Computer and Knowledge Engineering

An Effective Connectomics Approach for Diagnosing ADHD using Eyes-open Resting-state MEG

Authors :

Nastaran Hamedi¹ Ali Khadem² Sajjad Vardast³ Mehdi Delrobaei⁴ Abbas Babajani-Feremi⁵

1- Department of Biomedical Engineering. Faculty of Electrical Engineering. K. N. Toosi University of Technology. Tehran, Iran 2- Department of Biomedical Engineering. Faculty of Electrical Engineering. K. N. Toosi University of Technology. Tehran, Iran 3- Department of Electrical Engineering. Sharif University of Technology. Tehran, Iran 4- Department of Biomedical Engineering. Faculty of Electrical Engineering. K. N. Toosi University of Technology. Tehran, Iran 5- Dell Medical School University of Texas at Austin. Austin, TX, USA

Keywords :

Attention Deficit Hyperactivity Disorder (ADHD), Resting-state Magnetoencephalography (rs-MEG), Effective connectivity, Granger Causality (GC), Machine learning, minimum redundancy maximum relevance (mRMR) feature selection

Abstract :

Attention Deficit Hyperactivity Disorder (ADHD) is the most common neurological disorder in childhood. The apparent symptoms of this disorder include a significant lack of attention, impulsivity, and hyperactivity. As this disorder causes school problems and social incompatibility, an accurate diagnosis can help to mitigate these problems. Magnetoencephalography (MEG) is an appealing technique for diagnosing ADHD due to being non-invasive and having a high temporal and a good spatial resolution. In this paper, we proposed an effective brain connectomics approach based on eyes-open resting-state MEG (rs-MEG) to identify subjects with ADHD from healthy controls (HC). We calculated the effective connectivity between the MEG sensors using Granger causality (GC). Then we used those GC measures as input features of six base classifiers and a meta-classifier based on stacking ensemble learning using those six base classifiers to classify ADHD and HC. The most discriminative features selected by the minimum redundancy maximum relevance (mRMR) feature selection algorithm were fed to all base classifiers and also the meta-classifier. We achieved an accuracy of 89.82% using ten GC features and the meta-classifier. Moreover, our proposed method outperformed eyes-closed rs-MEG studies in the diagnosis of ADHD. It is noteworthy that there has been no study on diagnosing ADHD using either eyes-open rs-MEG or GC. Thus, the novelty of our proposed method is to use eyes-open rs-MEG data and GC to diagnose ADHD. Our results demonstrate the capability of an effective connectomics approach based on eyes-open rs-MEG for diagnosis of ADHD.