Artificial intelligence can diagnose autism in children by eye photos with an accuracy of 100%.
CTOnews.com, December 21 (Xinhua) researchers at the School of Medicine of Yonsei University in South Korea recently made a breakthrough: the use of deep learning artificial intelligence algorithm to analyze children's retinal photos unexpectedly achieved the diagnosis of autism spectrum disorder (ASD) with 100% accuracy.
Pixabay autism spectrum disorder is a neurodevelopmental disorder that children usually exhibit abnormalities in social communication, repetitive behavior and interest. Early diagnosis is very important for intervention treatment, but at present, the diagnosis of autism mainly depends on behavioral observation and professional evaluation, which is time-consuming and influenced by subjective factors.
This study opens up new possibilities for the diagnosis of autism. The researchers recruited 958 participants with an average age of 7.8, half of whom were autistic and the other half a normal control group. They took pictures of the retina of all participants and analyzed the association between the photos and the symptom severity score using a deep learning algorithm.
Surprisingly, the algorithm showed amazing accuracy in the test phase, successfully identifying all autistic children with an accuracy of 100%. Even if 95% of the least important regions in the image are removed (only the optic disc region is retained), the accuracy of the algorithm remains stable, indicating that the optic disc region is very important for distinguishing autism.
CTOnews.com noted that in addition to diagnostic accuracy, the algorithm can also assess the severity of autism symptoms, with an average accuracy of 0.74. The researchers point out that this number is in the "acceptable" range, with 0.8 to 0.9 being "excellent".
It is worth mentioning that the children involved in the study were as young as four years old, suggesting that the AI-based model could be used as an objective screening tool from this age group. Since the retina of a newborn does not stop developing until the age of four, further research is needed to determine whether the tool is suitable for younger children.
The findings, recently published in the prestigious medical journal JAMA Network Open, bring new hope to the field of diagnosis and treatment of autism.
