Research Progress in the Application of Deep Learning for the Diagnosis of Autism in Children

Autism Spectrum Disorder (ASD) is also known as autism. Since its first identification in 1943, the number of individuals with autism has continued to rise, and research on auxiliary diagnostics for autism has become a hot topic in recent years. Currently, the researched auxiliary diagnostic methods for ASD encompass the following three approaches: Firstly, utilizing the ABIDE dataset to analyze brain imaging of children with ASD, such as MRI and brainwaves, to identify differences between typically developing children and those with autism in terms of brain imaging or brainwaves. Secondly, studying the facial expressions of children with ASD and employing deep learning models like CNN to train and validate facial expression datasets; subsequently, using the model's predictions and classifications of unknown samples to determine whether a child has autism and the severity of their symptoms. Thirdly, gaze behavior is another avenue, where eye trackers or depth cameras are used to record eye movement data of children with ASD and typically developing (TD) children, and analyze the differences in their gaze patterns for diagnosis.