Deep learning AI models could be used to screen for autism and check on the severity of the condition, according to new research – and all the AI might need is a photo of the subject's retina. Previous studies have linked changes in retinal nerves with altered brain structures, and from there to Autism Spectrum Disorder (ASD). The evidence suggests the eye really is a window to the brain, via the interconnectedness of the central nervous system. In a new study, a team from the Yonsei University College of Medicine in the Republic of Korea wanted to see if AI could spot ASD in retinal patterns. First, the model was trained on images where the AI was told if the subject had autism or not.
Then, the AI was asked to analyze the retinas of 958 children and teenagers with an average age of 7.8 years overall, half of whom had been diagnosed with autism. It managed a perfect score in identifying those who had autism and those who didn't. The AI wasn't quite as good at predicting symptom severity from the retina photos (was only accurate between 48-66 percent of the time). Still, there's lots of potential here to help kids get a useful assessment at an earlier age. In this case, the researchers limited the training data to children and adolescents between the ages of 4 and 18, but they think that it could work on even younger kids as well – that's something that future studies might be able to look into.
More information:
https://www.sciencealert.com/a-hidden-pattern-in-childrens-eyes-can-reveal-if-they-have-autism