ML Based Approach to Detect Autism Spectrum Disorder (ASD)

Abstract

Autism Spectrum Disorder is an exponential disorder which causes notable challenges in community interaction and behavior for the people affected. Autism children may have trouble in interacting with others and learning the meaning of words for them is really difficult. Parents observe strange behaviors from autism children like lack of coordination and repetition of physical movements such as rotating, waving their hands in a rapid motion, violent shaking of head and body. Autism Spectrum Disorder (ASD) cannot be cured but makes a huge difference in the lives of many children if early treatment is given. And so, ASD is termed as lifelong disorder. Based on the complexity of the disorder, symptoms and severity, the cause for ASD differs. Genetics and environment play a major role. Children having fragile X syndrome and other genetic diseases or children of older parents or exposure to environmental toxins tend to develop autism. In China, 1 out 9000 kids and in India, 23 out of 10000 kids are affected. Early detection is required to improve better treatment methodology and enhance the quality of life of ASD suffering people. Unfortunately, there is no special test to detect Autism. Autism disorder is usually detected by observing the behavioural activities of the children. This method of diagnosis is time consuming and not suitable for early detection. This paper primarily focuses on using ML model to diagnose Autism at an early stage. ML models are usually working with the relationship among various brain regions and hence are preferred over other methods. Moreover, machine learning algorithms are characterized with more prediction accuracy over other models. So, ML which is a subfield of Artificial Intelligence, can be used to enhance the detection methods of Autism by exploring its genetics and designing effective interventions.