Transfer Learning Approach for Detection of Autism Spectrum Disorder using Facial Images

Autism Spectrum Disorder (ASD) is characterized by deficits in social interaction, communication issues, and repetitive behavioral patterns. Effective ASD intervention and assistance depend on early detection and diagnosis. Children with autism differ noticeably from normally developing (TD) children in their patterns of facial features. Autism may be indicated by a child's eyes, nose, and lip distances as well as their placement. Using common evaluation measures like accuracy, sensitivity, and specificity, the built ASD detection system is assessed. The outcomes of this study show the utility of facial image analysis in the identification of ASDs. The proposed strategy shows potential in aiding physicians and researchers in early diagnosis by utilizing face clues and machine learning techniques. This study aims to investigate how well machine learning and feature extraction algorithms based on deep learning can identify ASD in young children. Deep convolutional neural networks known for their capacity to extract significant characteristics from images, VGG16 and VGG19, are used. Several machine learning algorithms, such as logistic regression, support vector machines (SVM), naive Bayes,, and artificial neural networks (ANN), are then applied with the retrieved characteristics as input.