A new approach to diagnosing autism

Researchers in America have succeeded in developing machine learning models to predict the likelihood of children being diagnosed with autism .

The study was conducted by researchers from the Penn State research team in Pennsylvania, and was published in BMJ Health & Care Informatics , and reported by EurekAlert.

This approach may assist clinicians in diagnosing the disease, and it may be appropriate to combine it with traditional methods of diagnosing autism.

Increasing cases of autism and the importance of early diagnosis Autism spectrum disorder begins with the sufferer from birth and appears during the first 3 years and continues throughout life and is known as a neurological deficiency and complex developmental deficiency.  19.02.2022

Building models

The researchers developed a model that correlates hundreds of available clinical variables about the child, including visits to the doctor and medical services provided to him in cases that may appear unrelated to autism.

The researchers used the data available from insurance companies to build their models, and inserted the available data into several machine learning systems to train them to evaluate variables and find relationships that may increase the likelihood of developing autism spectrum disorders.

Dr. Koshi Shen, an assistant professor at Penn State University's College of Engineering, who was co-author of the study, says that information from insurance companies about patients is abundant, and when used, it gives comprehensive information about people's medical history.

Dr. Koshi Shin adds that, according to previously published research, the medical history of children diagnosed with autism may include distinct conditions, including certain types of infections, epileptic seizures, digestive disorders and some distinctive behaviors.

Calculating the likelihood of developing autism

These conditions do not cause autism, but may be a characteristic of children with autism, which has inspired researchers to collect patient data and try to link it and use it to predict and calculate the likelihood of developing autism.

Dr. Judong Lu, a specialist in mental health and behavioral health at Penn State University School of Medicine, co-author of the study, comments that the doctor needs to take several notes and visits to diagnose the child, which requires time, and may cause the child to miss out on early medical intervention that gives him the opportunity to achieve the best response.

Dr. Shen says that the model they built combines factors that increase a child's likelihood of developing autism, and gives a ratio for that probability.

These predictions are very close and possibly better than the predictions given by traditional tools used in diagnosing the disease. Dr. Shen believes that the combination of traditional diagnostic tools and these models may be a promising approach to assist clinicians with diagnosis.

Dr. Lu adds that it is feasible and feasible to combine traditional diagnostic methods with these models.

What is autism?

According to the World Health Organization , “Autism spectrum disorders are a group of diverse disorders characterized by some difficulties in social interaction and communication. Other features of these disorders are atypical patterns of activities and behaviors, such as difficulty moving from one activity to another, dwelling on details, and abnormal reactions Ordinary feelings.

"The abilities and needs of people with autism vary, and can develop over time. Some people with autism may be able to lead independent lives, but others have severe disabilities and require lifelong care and support. Autism often affects education and employment opportunities." In addition, the burden of care and support for families may increase. Community behaviors and levels of support from local and national authorities are important factors that determine the quality of life for people with autism."

Note that about one in every 100 children suffers from autism.