Analysis of Brain Imaging Data for the Detection of Early Age Autism Spectrum Disorder Using Transfer Learning Approaches for Internet of Things

In recent years, advanced magnetic resonance imaging (MRI) methods including functional magnetic resonance imaging (fMRI) and structural magnetic resonance imaging (sMRI) have indicated an increase in the prevalence of neuropsychiatric disorders such as autism spectrum disorder (ASD), effects one out of six children worldwide. Data driven techniques along with medical image analysis techniques, such as computer-assisted diagnosis (CAD), benefiting from deep learning. With the use of artificial intelligence (AI) and IoT-based intelligent approaches, it would be convenient to support autistic children to adopt the new atmospheres. In this paper, we classify and represent learning tasks of the most powerful deep learning network such as convolution neural network (CNN) and transfer learning algorithm on a combination of data from autism brain imaging data exchange (ABIDE I and ABIDE II) datasets. Due to their four-dimensional nature (three spatial dimensions and one temporal dimension), the resting state-fMRI (rs-fMRI) data can be used to develop diagnostic biomarkers for brain dysfunction. ABIDE is a collaboration of global scientists, where ABIDE-I and ABIDE-II consists of 1112 rs-fMRI datasets from 573 typical control (TC) and 539 autism individuals, and 1114 rs-fMRI from 521 autism and 593 typical control individuals respectively, which were collected from 17 different sites. Our proposed optimized version of CNN achieved 81.56% accuracy. This outperforms prior conventional approaches presented only on the ABIDE I datasets.