Autism detection from facial Images using deep learning methods

Autism spectrum disorder (ASD) refers to a collection of behavioral and developmental issues and difficulties. The cognitive, communication, and play skills of a child with autism spectrum disorder are all affected. The description "spectrum" in autism spectrum disorder refers to the fact that each child is special and has their own set of characteristics different from other children. These come together to give him a unique social bond as well as his own understanding of his own actions. Medical image classification is a significant research field that is gaining traction among researchers and clinicians alike to detect and diagnose diseases. It addresses the issue of medical diagnosis, experiment purposes and analysis in the field of medicine. Several deep learning-based medical imaging applications have been proposed and developed to understand and learn about how diseases develop in patients, to help doctors in early diagnosis of pathology. Not only is achieving good accuracy in classifying medical images the main purpose alone. We used pre- trained CNN and transfer learning in this study. These CNNs pre-trained architectures are used to train the network and to classify medical images. The experimental results of this study show that the model based on transfer learning (Support Vector Machine + MobileNet) can detect Autism Spectrum Disorder at the best rate with 98.83% accuracy. The architectures that were tested in this study are ready to be tested with additional data and can be used to prescreen individuals with ASD. The use of deep learning methods for feature selection and classification in this study could greatly support future autism studies.