Autism Spectrum Disorder Detection using theDeep Learning Approaches

Abstract

This Autism Disorder is a developmental impairmentthat affects how a person perceives, communicates, and behaves.It is brought on by changes in the brain. Before the age of three,ASD develops, and it can persist up to death. The self-harm attempts made by those who have this disorder are significantly higher than people without have it. For diagnosing this disorder at an early stage, early detection is necessary. Various machine learning (ML) techniques namely- Random Forest, SVM, NaiveBayes, Decision Tree, etc., and deep learning (DL) approaches such as VGG16, DenseNet, AlexNet, etc., can be utilized on the questionnaires filled during the conducted survey based on behavior whereas, images can also be provided as input to these approaches for identifying of ASD. The proposed methodology suggests the detection of autism using deep learning algorithms based on transfer learning. ASD identification model consists of various stages namely- data collection/acquisition, pre- processing, data augmentation, feature extraction, and the last stage is classification.