Autism spectrum disorder (ASD) is one of the most common diseases that cause difficulties for an individual to express his/her emotions or to understand other's emotions. ASD has become a challenging problem as its symptoms are unpredictable. The main symptoms of ASD include problems such as abnormal social reciprocity, nonverbal communication, sensory abnormalities, etc. To understand such abnormalities, there is a requirement of some learning tools. It has been witnessed that facial expression images, eye tracking, and neuroimage have been shown as effective tools for analysis of abnormalities that had occurred in both grey and white matter of the brain. Many researchers focused their work on the classification problem of ASD disorder from healthy subjects but still didn't reach effective diagnosis and healing tools. As with the advancement of digital image processing, it has become feasible to use these technologies for accurate diagnosis of ASD subjects. These technologies are integrated with deep learning for the identification and treatment of ASD.
Top1. Introduction
Autism can be categorized as the disorder of neurodevelopmental disorder which has symptoms like an early stage of shortcoming in social communication while interacting with people, being restricted and confined, and having a behavior of repetitive pattern. In the age group of fewer than five years, the main reason disabilities of mental disorders are autism. If the childhood disorders like; Attention Deficit Hyperactivity Disorder, as well as Conduct Disorder, is compared with autism then the effect of autism is for a lifetime. The symptoms of ASD can be recognized at its early stage but there are still some disabilities and pattern of behaviors are difficult to recognize as symptoms till the life of a child is affected at a significant level. Individuals with ASD would have varied limited functions or symptoms and it may differ over the period. One of the main symptoms of ASD is delayed or misinterpretation in language acquisition. It is reported that around 50 percent of people suffering from autism are incapable of framing a clear speech. ASD’s one of the basic symptoms is impairment in social communication. It can be observed when the babies or young children do not respond properly to physical contact or interaction. Along with these symptoms; some other patterns like repetitive behavior or having a repetitive interest or some unusual interest in useless engrossment or some wired interest can be listed as the indicators of ASD (Kumar et al., 2019). Mostly ASD symptoms can be spotted in a child up to the age of 18 months. But many times, ASD is not diagnosed accurately in case the child is suffering limited delayed speech and such cases are diagnosed when children face problems in interacting with friends (Chen et al., 2019). Various tools, as well as approaches to do the diagnosis of ASD, are applied by doctors in conjunction with diagnostic tools. Doctors use classification methods for these types of studies (Jiang et al., 2019). In short, it can be said that machine learning algorithms are being used widely for the diagnosis of ASD.
1.1 Motivation and Scope
The term “autism spectrum disorder” refers to a group of neurodevelopmental disorders. ASD is characterized by poor emotional control and social interactions, as well as restricted interest, repetitive behaviors, and sensory hypo-/hyperreactivity. Learning, development, control, and interaction, as well as some everyday life abilities, are frequently impaired in people with an autism spectrum disorder. ASD places a significant financial strain on patients’ families and society. Establishing an early and reliable diagnosis framework to distinguish ASD patients from usual controls is critical (TC). Neuroimaging techniques that have not recently been invasive and in vivo have become an area for intensive research on ASD auxiliary diagnoses. Neuroimaging technology is now widely employed in the investigation of a variety of brain illnesses, including ASD. Magnetic resonance imaging (MRI) provides high-grade three-dimensional (3D) images and detailed structural information on brain structures. The associated disorders have been subjected to morphological analyses based on MRI imaging, with promising outcomes. Therefore this chapter explores the application of image processing in the diagnosis of ASD.
1.2 Objectives
It is quite difficult to differentiate and recognize the person suffering from ASD and typical controls (TC) persons. In existing or conventional approaches, ASD/TC can be differentiated based on morphological features. So, this chapter is mainly designed to explore the benefits of image processing and deep learning to study extracted patterns of the brain from MRI or fMRI images. In this chapter, a model is also presented with features of CNN learning to perform ASD/TC classification.
1.3 Problems Statement
The main symptoms of ASD include problems such as abnormal social reciprocity, nonverbal communication, sensory abnormalities, etc. To understand such abnormalities there is the requirement of some learning tools. Many researchers focused their work on classification problem ASD disorder from healthy subjects but still didn’t reached effective diagnosis and healing tool. The existing ASD diagnosis tools are mostly based on behavior observation and symptoms which can be sometimes misdiagnosed. So, to develop a more qualitative diagnosis, there is a need for advanced tools (Reyana, & Kautish, 2021, (Rani, & Kautish, 2018) and artificial intelligence interface to visualize and analyze the biological abnormalities for ASD disorders