Efficacy of novel Summation-based Synergetic Artificial Neural Network in ADHD diagnosis

Abstract

Abstract Attention Deficit Hyperactivity Disorder (ADHD) is a critical condition that affects millions of children and often continues into adulthood. In this paper, we propose a dual 3D CNN data integration platform, that we call SSANN, a Summation-based Synergetic Artificial Neural Network, for ADHD diagnosis. The diagnosis problem in our research is simplified into a binary classification problem to detect an ADHD affected or a typical developing child given magnetic resonance imaging scans. Our proposed model has two 3D CNN branches with varying structures: (1) The first branch extracts features from the functional MRI (fMRI) data from the subjects, (2) the second branch extracts features from the structural MRI (sMRI) data of the corresponding subjects. Later, output matrices of both branches are combined with a proposed summation induced process which then is fed into a fully connected neural network and finally produce the binary classification prediction. Our proposed model achieved accuracy of 72.89% on the ADHD-200 dataset, which performs superior than the state-of-the-art approaches on the same evaluation dataset. Our proposed SSANN model can extract useful features from the fMRI and sMRI data, which can also be employed to understand the brain anatomy and its functions in the subjects better, as well as can be used in other machine learning algorithms to design and develop diagnostic tools that certainly has potential to escort ADHD research move forward. Our proposed model offers a robust multi-modal data integration platform that can also be adapted in other medical imaging domains.