Children's Eyes May Contain Hints of Autism Spectrum Disorder; Study Reveals Retinal Photographs Could Be Viable For Screening Procedures

A new study reveals that deep learning models of AI could be used for autism screening and severity assessment. What's more is that all it may take for the AI to make the analysis is retinal photographs.

Retinal Changes and Autism Spectrum Disorder

In the past, retinal nerve changes have been linked to autism spectrum disorder (ASD). The evidence posits that the eye is truly a window to see the brain through the central nervous system's interconnectedness.

Previously, researchers were also able to connect the light response of the retina to cases of ASD and attention deficit hyperactivity disorder (ADHD). This also shows how eyes can shed light on the brain activity of an individual.

In the new study entitled "Development of Deep Ensembles to Screen for Autism and Symptom Severity Using Retinal Photographs," the researchers note that individuals who have autism experience structural changes in the retina that could possibly reflect changes in the brain. This includes abnormalities in the visual pathway via anatomic and embryonic connections. They add that if deep learning algorithms could be utilized in objective ASD screening or severity assessment through retinal photographs remains a mystery.

Using AI To Screen For Autism, Severity Through Retinal Photographs

As part of the new study, the researchers wanted to see if AI was capable of detecting specific patterns in the retina. With this, they first trained the AI model with various images where the AI was notified if the subject did or did not have autism.

Moving forward, the AI was then tasked to analyze 958 children and teenagers' retinas. These participants had a mean age of 7.8 years. Half of them received an autism diagnosis. The AI was able to identify those with and without autism with a perfect score.

However, when it comes to symptom severity prediction, the AI was not very good with it. The model only boasted of an accuracy that ranged between 48% to 66%. Nevertheless, this shows great potential for offering useful assessments to kids at a younger age.

For the study, the training data used was limited to adolescents and children from 4 to 18 years old. The researchers do believe that it could also work for children who are even younger, which is something that further studies could possibly examine.

Roughly one in 36 people is believed to be dealing with ASD. Being aware of ASD as early as possible could result in great differences for people to understand themselves and navigate through the world.

The study findings suggest that using retinal photographs is a possibility for objectively screening for autism and even assessing severity of symptoms.

Check out more news and information on Medicine & Health in Science Times.