A 17-year-old New Jersey student won a $175,000 science prize for training AI to detect signs of autism and ADHD from retinal images - a project that could point toward faster screening for conditions that often take months or years to diagnose.
Edward Kang, a student at Bergen County Academies in Hackensack, developed RetinaMind, a project that combines artificial intelligence with laboratory biology to investigate whether photographs of the eye can reveal patterns associated with neurodevelopmental disorders. The prize, awarded through the Society for Science, recognizes Kang's work as one example of how young researchers are applying AI to complex medical problems.
The eye as a window into brain development
The retina is derived from the same tissue as the brain, and previous research has observed differences in the retinas of people with certain neurodevelopmental conditions. Kang's project tested whether AI could detect those subtle differences in retinal photos and potentially serve as a screening aid.
He trained AI models on a large public dataset of retinal images, using several techniques to improve performance. He also examined which features influenced the model's predictions - a critical step in healthcare, where researchers need to confirm that an AI system is detecting a real biological signal rather than an unrelated pattern.
Kang then built a prototype called RetinaMind to demonstrate how such a screening tool could operate.
Bridging computational models and cell biology
The project extended beyond computer models. Kang developed a model of a retinal cell to investigate gene changes that might explain the patterns observed in the images, then confirmed his findings with a second cell model. That combination of computational analysis and laboratory experiments adds biological evidence to what the AI detected.
Kang's work touches on a growing area of research: using AI for Science & Research to identify patterns that might be invisible to the human eye. The approach is not a clinical diagnostic tool, but it demonstrates a possible pathway for earlier screening of conditions like autism and ADHD.
His broader profile includes a runner-up finish in the International Research Olympiad, co-founding a club that teaches students how to read scientific research, and serving as safety co-chair and judge for his middle school science fair. He also sings baritone in the New Jersey All-State Mixed Chorus and has performed at Carnegie Hall.
Why this matters for science and research professionals
For researchers working in AI for Healthcare, Kang's project demonstrates a useful pattern: pairing AI model training with laboratory validation. Many machine learning projects stop at prediction accuracy, leaving open the question of whether the model is learning something biologically meaningful. Kang's approach - investigating gene changes and confirming findings across two cell models - offers a template for strengthening the credibility of AI-based screening research.
Your membership also unlocks:
