New Aid for Early Diagnosis in Autism: Artificial Intelligence

Advances in artificial intelligence technologies are opening up new opportunities in the early screening, assessment and intervention processes of Autism Spectrum Disorder (ASD). The ability of AI-based systems to analyze patterns in different data such as eye movements, tone of voice, facial expressions, attention processes and play behaviors is accelerating research aimed at identifying risk indicators at an early stage.

Üsküdar Üniversitesi Sağlık Bilimleri Fakültesi Çocuk Gelişimi Bölümü Dr. Öğr. Üyesi Demet Gülaldı assessed developments at the intersection of artificial intelligence technologies and Autism Spectrum Disorder.

Noting that artificial intelligence has transformed from being merely a data-processing technology into systems capable of analyzing complex patterns in human behavior, Gülaldı stated that developments particularly in machine learning, deep learning and natural language processing offer new possibilities for autism research.

Artificial intelligence can analyze patterns in human behavior
Noting that studies on artificial intelligence and autism have gained momentum in recent years, Dr. Öğr. Üyesi Demet Gülaldı said that the technology's ability to analyze large amounts of data in a short time provides a significant advantage for research.

Gülaldı said, "It is no longer possible to view artificial intelligence merely as a data-processing system. Today, these systems can learn, analyze human behaviors, and solve very complex patterns in a short time. In particular, developments in the fields of Deep Learning and Natural Language Processing (NLP) allow machines to go beyond simply carrying out commands and analyze nuances in human behavior."

Behavioral indicators play an important role in autism diagnosis
Stating that Autism Spectrum Disorder is a neurodevelopmental condition characterized by differences in social communication and interaction as well as restricted and repetitive behavior patterns, Gülaldı noted that clinical assessments hold an important place in the diagnosis process.

Pointing out that the causes of autism are multifaceted, Gülaldı said that genetic and environmental factors play a role, but due to differences among individuals, clinical assessments today are largely based on observable and measurable behavioral indicators.

Early intervention is critical for development
Stating that early diagnosis and intervention in autism are of critical importance for the child's developmental process, Gülaldı noted that although symptoms mostly emerge in early childhood, the diagnosis process can be delayed to later ages in some children.

Gülaldı said, "Early diagnosis and intervention can contribute to reducing difficulties in social attention, language and cognitive development areas in individuals with autism, as well as lowering the overall severity of symptoms. For this reason, it is of great importance to notice risk indicators at an early stage and direct individuals to appropriate assessment processes."

Different data ranging from eye movement to tone of voice can be evaluated
Stating that the combination of artificial intelligence and autism research offers remarkable opportunities especially in the field of early screening, Gülaldı said that different behavioral data can be evaluated together through algorithms.

Gülaldı used the following words: "Machine learning algorithms can analyze patterns in data obtained from a child's eye movements, frequency changes in tone of voice, or the way they play. These studies constitute an important research area aimed at noticing certain signs associated with autism at an earlier stage."

Noting that AI-supported intervention tools can also contribute to the development of personalized education and support processes tailored to children's individual characteristics, Gülaldı emphasized the importance of treating the technology as an aid that supports assessment and intervention processes rather than as a tool that replaces experts.

Machine learning can accelerate diagnosis processes
Reminding that information obtained from parents and clinical observations play an important role in traditional early screening processes, Gülaldı said that machine learning-based systems can support assessment processes by making use of large data sets.

Gülaldı continued as follows:

"Traditional methods used in the early screening of children with autism proceed with psychiatrists assessing autism symptoms in the child's development based on parental statements and clinical observations. Artificial intelligence methods such as machine learning, on the other hand, can support and accelerate the assessment process by learning patterns in data obtained from a large number of individuals with autism and control groups. For example, promising results are reported for Convolutional Neural Network (CNN)-based ASD classification systems in terms of performance metrics such as accuracy, sensitivity and specificity."

AI-based models are being researched in young children
Noting that there are also studies on the use of AI-based models for early assessment in young children, Gülaldı stated that high accuracy rates have been reported in some studies.

Gülaldı said, "In one study, it was reported that an alternating decision tree model created with artificial intelligence achieved a result with 99.97 percent accuracy in autistic children aged 13 to 48 months. Such studies demonstrate the potential of artificial intelligence for identifying behavior patterns associated with autism, particularly during early childhood."

However, before such research results are directly transferred to clinical practice, they need to be validated in different groups, supported by independent studies, and considered together with the assessments of health professionals.

Personalized intervention processes may become possible
Stating that the use of artificial intelligence is not limited to early screening and assessment alone, Gülaldı said that technology-supported systems can also contribute to personalized applications in intervention and education processes.

Gülaldı concluded by saying, "AI-supported systems can analyze children's behaviors and evaluate eye contact, facial expressions and attention processes. The development of these systems can contribute to the development of education and intervention programs adapted to children's individual characteristics. Artificial intelligence is now becoming an important research and technology field that can support access to early diagnosis and the right support in the field of autism."