Deep Learning Approach for an early stage detection of Neurodevelopmental Disorders

Neurodevelopmental disorders are highly heterogeneous disorders. The symptoms are not same amongst all the individuals and cannot be detected easily by looking at the physiological changes in the individuals. The cause of these disorders may be genetics related but exact causes are not known till date. These disorders can last through one's life if proper treatment is not provided at early stages. Neurodevel-opmental disorders mainly include Autism, ADHD, Schizophrenia. Neurodevelopmental disorders can be identified by using sophisticated technologies. The Functional Magnetic Resonance Imaging(fMRI) is preferred to identify the neurodevelopmental disorders since it allows to measures the activity of brain by detecting changes associated with the blood flow. In this paper, we present a deep learning based system developed to detect the Autism, ADHD, Schizophrenia disorders. The proposed system is trained using ABIDE, ADHD 200, COBRE, UCLA, WUSTL datasets. It is observed that the proposed system is able to produce the results with 71.16% accuracy, 70.13% precision, 69% sensitivity, 80.80% specificity, and 69.56% F1-score.