AI models using retinal images achieve perfect accuracy in diagnosing autism

Researchers developed and tested deep ensemble models (DEMs) to automatically diagnose autism spectrum disorder (ASD) and determine symptom severity through high-resolution retinal images in a study published in JAMA Network Open. They used 1,890 photos from 958 participants and found their models could diagnose ASD with a high level of accuracy. Additionally, the models could estimate ASD severity with significant accuracy, even in children as young as four years old. This research demonstrates the potential for computer-aided diagnosis of ASD using retinal photographs, which could address the lack of expert resources for diagnosis and improve early and broad screening for ASD.

Source link