A Review of Machine Learning Techniques for Detecting Autism Spectrum Disorders

Social difficulties, monotonous behavioral trends, speech, and nonverbal cues are all characteristics of autism spectrum disorder (ASD). Autism spectrum disease affects one out of every 160 children world - wide, on the report of WHO, and in India, it estimates to be one out of 500. They would benefit greatly from early identification during their first few years on earth; however, acknowledging and diagnosing ASD is frequently delayed. As a result, developing a detection model relying on functional or structural brain region interactions, such as ASD, is vital for diagnosis. This assists doctors in making a diagnosis by allowing them to examine the child's developmental history and behavior. Here, the primary objective is to identify existing ASD detection methods and summarize their similarities and dissimilarities, culminating in tentative hypotheses that can be affirmed by proper research. This paper presents a comparison of state-of-the-art methods in the literature. The paper also discusses various challenges in autism detection and also research opportunities in the area of autism spectrum disorder detection.