Intelligent Mobile Application for Autism Detection and Level Identification System Using Deep-Learning Model

The early detection and level assignment are very important in autism spectrum disorder (ASD) cases.In this paper, we introduce an autism diagnosing and level identification mobile application based on a deep-learning model.The application is designed with two stages; the first stage classifies children as either ASD children or potentially normal, while the second stage identifies the ASD level.For feature extraction and image classification, a convolutional neural network (CNN) is proposed.The 2,122 photos that were removed from the 3000 original dataset because of poor quality and racial imbalances made up the dataset used to develop and evaluate this model.Results show that the accuracy is 97.3% and the area under the curve is 99.8%.The identification of ASD level is performed using two welldefined scales in the literature that are adopted in the traditional examinations.Depending on the child's ASD level, psychiatrists provide specific and standard learning support.The application provides caregivers with fast decisions (about 20 minutes) to get an idea of the learning strategy to be followed with their child.Early intervention is very beneficial for children diagnosed with ASD, and these applications assist children in underfunded nations or those without access to healthcare.