A New Horizon in ADHD Screening: AI + Eye Imaging
Attention-deficit/hyperactivity disorder (ADHD) remains one of the most common neurodevelopmental conditions in children, yet diagnosing it reliably—with speed, objectivity, and scalability—has proven challenging. The diagnostic pathway conventionally relies heavily on behavioral evaluation, parent/teacher reports, and neuropsychological testing. These methods are valuable, but they bring subjectivity and resource burdens.
In March 2025, researchers led by Hangnyoung Choi published a study titled “Retinal fundus imaging as biomarker for ADHD using machine learning for screening and visual attention stratification”
in npj Digital Medicine, a Nature Medicine partner journal.1
The study presents a noninvasive retinal imaging technique enhanced by artificial intelligence (AI) to identify ADHD biomarkers and stratify visual-attention-related executive function (EF) subdomains.
Study Overview
The study analyzed 1,108 retinal fundus photographs from 648 children (mean age 9.5 years), including 323 with clinically diagnosed ADHD and a matched control group.2
Using the AutoMorph image-processing pipeline, the researchers extracted structural retinal features—vessel density, vessel morphology, and optic disc shape—and then trained four machine-learning classifiers (including XGBoost and Random Forest) to:
- Distinguish ADHD from typical development
- Predict EF subdomain performance, particularly in visual attention tasks
Major Findings
- Exceptional accuracy: The AI models achieved AUROC scores between 95.5 and 96.9 percent, indicating strong discriminative power.1
- Strongest indicators: Retinal vessel density and optic disc shape were the most consistent and influential predictors of ADHD.3
- Executive function correlation: Visual-attention EF subdomain scores correlated strongly with retinal features (AUROC > 85%), while auditory-related EF performance showed weaker associations.4
- Neurodevelopmental link: The researchers propose that retinal vascular changes mirror cerebrovascular and neurodevelopmental alterations found in ADHD, providing a direct anatomical window into the brain’s neural circuitry.5
Clinical Significance
This Harvard-affiliated study highlights the potential of retinal imaging as a fast, objective, and radiation-free screening method—especially valuable for pediatric evaluation, where behavioral assessments can be inconsistent. The approach could:- Reduce reliance on subjective behavioral testing
- Identify ADHD subtypes through executive-function stratification
- Allow clinicians to monitor treatment responses, as retinal structure may reflect dopaminergic therapy effects (e.g., methylphenidate-related changes in retinal thickness or microglial activity).6
Limitations and Future Directions
- Population bias: The data originated from South Korean cohorts; replication across diverse ethnicities and global populations is essential.2
- Correlational nature: The study cannot yet determine whether retinal alterations are causative or secondary markers.
- Implementation challenges: Clinical adoption requires standardization in imaging, training datasets, and cost-effectiveness analysis.
- Refinement opportunities: Combining fundus photography with optical coherence tomography angiography (OCT-A) could improve biomarker precision and resolution.7
Conclusion
This groundbreaking research demonstrates how AI analysis of retinal fundus images can provide an objective biomarker for ADHD, offering insight into both visual-attention mechanisms and underlying neural circuitry. While further validation is necessary, the findings open the door to more precise, accessible, and early ADHD screening tools that integrate neurology, ophthalmology, and digital medicine.References
- Choi H., Hong J. S., Kang H. G., Park M. H., Ha S., Lee J., Yoon S., Kim D., Park Y. R., Cheon K. A. Retinal fundus imaging as biomarker for ADHD using machine learning for screening and visual attention stratification. npj Digital Medicine. 2025 Mar 17; 8(1):164. DOI: 10.1038/s41746-025-01547-9
- PubMed: 40097590
- ScienceAlert. “Something in the back of your eye could reveal whether you have ADHD.”
- SSRN Preprint: Choi H. et al., Retinal Fundus Imaging as a Biomarker for ADHD (2024)
- Lifeboat Foundation Blog. “Retinal fundus imaging as biomarker for ADHD using machine learning for screening and visual attention stratification.”
- Catholic University of Korea — Elsevier Pure entry
- OUCI Database Entry
