Why look for signs of autism in the eye?
Autism spectrum disorder affects communication, social interaction, and interests. In 2020, the CDC estimated that 1 in 36 children in the US had autism. Diagnosis relies on evaluations by trained specialists, and the authors point out that there are not enough of them, so wait times keep growing.
The retina, the thin layer at the back of the eye that senses light, grows from the same tissue as the brain. Earlier studies had found small differences in the retina of people with autism. So the researchers asked whether an AI could spot them on a simple photo of the back of the eye.
How was the study done?
The study took place at Severance Hospital in South Korea. It included 958 young people under 19, with an average age of 7.8: 479 diagnosed with autism by specialists, and 479 typically developing peers of the same age and sex. Photos of the children with autism were taken in 2022 for the study, while the others came from the eye clinic's records.
The team trained a group of five neural networks, programs that learn from examples, then tested them on photos held back. Results are given as the area under the curve: 0.5 means chance, 1 means a perfect sort.
What the AI found in the photos
To separate the two groups, the AI scored 1.00 on the test set: a perfect sort. It kept that score using only 10% of the image, the area around the optic disc, where the optic nerve leaves the eye. The result also held in the youngest children, from age 4.
To tell severe autism from mild to moderate autism, as rated by a specialist, the score was only 0.74. When severity came from a questionnaire filled in by parents, the AI did not work at all.
Key figures
- 958young people included (1,890 eyes), at a single hospital in Seoul
- 1.00area under the curve for separating autism from typical development on the test set
- 0.74area under the curve only, for rating symptom severity
While you wait for an evaluation, with Mima
This is still research: today, autism is diagnosed through an evaluation by professionals, and families can wait a long time for one.
If you are waiting for an appointment for your child, Mima helps you jot down what you notice day by day, no spreadsheet needed, and gather your questions for that visit, such as: "What can we do while we wait for the evaluation?" She also explains the words in the report in plain language.
What this might change, someday
A photo of the back of the eye is quick, painless, and does not need an autism specialist. The authors see a possible way to help flag children who should get a full evaluation sooner, in places where appointments are hard to get.
But today, no eye exam is used to diagnose autism. Diagnosis is still made by professionals, based on observing the child and talking with the family.
Why caution is needed
Everything happened in a single hospital, and a perfect score needs to be checked elsewhere. Photos of the two groups were not taken at the same time or in the same setting. The AI only compared children with autism to children without any condition: it did not learn to tell autism apart from other conditions, such as ADHD.
Children with other conditions were left out, even though many children with autism have them in real life. The authors call for studies across several centers, with younger children and more detailed retinal scans.
Frequently asked questions
Can a retinal photo diagnose autism?
No. This study is an early research lead from a single hospital. Autism is still diagnosed by trained professionals.
Why would the retina be linked to autism?
The retina grows from the same tissue as the brain and stays connected to it through the optic nerve. Earlier studies had found retinal differences in people with autism.
What ages were tested in the autism eye study?
The youngest children in the study were 4 years old. The AI was not tested in younger children.
Should my child get an eye exam to check for autism?
No, eye exams are not used to diagnose autism. If you have questions about your child's development, talk to your pediatrician.
The original study
Kim JH, Hong J, Choi H, et al. “Development of Deep Ensembles to Screen for Autism and Symptom Severity Using Retinal Photographs” JAMA Network Open, December 1, 2023. DOI : 10.1001/jamanetworkopen.2023.47692
Read the full study in JAMA Network OpenThis article summarizes a scientific study and is not medical advice.