By Sara Gazla

They say the eyes are the windows to the soul, but science suggests they may also offer a window into the brain.

The retina is an anatomical extension of the central nervous system (CNS), sharing the same embryological origins as the brain. This connection makes it a particularly interesting place to investigate neurological disease. Researchers have already identified changes in retinal structure in people with neurodegenerative conditions such as Alzheimer's and Parkinson's disease (London et al., 2013), suggesting that the eyes may reveal clues about what is happening within the brain.

These alterations could help researchers better understand the mechanisms driving neurodegeneration and potentially uncover new targets for treatment. Interestingly, retinal abnormalities have also been observed in individuals with mild cognitive impairment, suggesting that changes in the eye may be detectable during earlier stages of disease (London et al., 2013). This raises an intriguing question: could the eyes reveal signs of neurodegeneration before the first clinical symptoms appear? Although promising, further research is needed to establish whether retinal changes truly precede neurological symptoms.

What makes this possibility particularly significant is how easily the retina can be examined. Unlike the brain, which often requires complex and expensive imaging techniques, the retina can be visualised using non-invasive methods such as optical coherence tomography (OCT), which is already widely used in clinical practice. The retina's organised neuronal layers allow researchers to measure subtle structural changes, generating valuable data that artificial intelligence (AI) could analyse to identify patterns associated with neurological disease.

The potential of AI-assisted retinal imaging is already being demonstrated in clinical research. One example is the detection of papilloedema, swelling of the optic disc that can indicate dangerously raised pressure within the skull. The BONSAI deep-learning system correctly identified 96.4% of cases with papilloedema while accurately ruling out the condition in 84.7% of cases without it (Sathianvichitr et al., 2023). This highlights how AI could help clinicians recognise important neurological warning signs through a simple retinal photograph.

But the possibilities extend beyond detecting visible abnormalities. In a 2025 study, Choi et al. used machine-learning algorithms to distinguish children with attention-deficit/hyperactivity disorder (ADHD) from typically developing children using retinal photographs. The model achieved an AUROC of up to 0.969, indicating strong discrimination between the groups. Features including retinal vessel density, vessel width and optic disc shape contributed to its predictions, demonstrating how AI can recognise subtle patterns associated with neurodevelopmental differences.

Similar approaches are being explored for Alzheimer's disease. A large retrospective study involving 12,949 retinal photographs found that deep-learning algorithms could distinguish patients with Alzheimer's dementia from controls, achieving accuracies ranging from 79.6% to 92.1% across independent cohorts (Sathianvichitr et al., 2023).

Despite these findings, challenges remain. Similar retinal abnormalities occur across different neurological conditions, making it difficult to determine whether AI is identifying disease-specific changes or broader patterns of neurological damage. Larger, more diverse studies following patients over time are therefore essential before retinal imaging can become a reliable neurological screening tool.

One example of where this technology could be heading is RetinaMind, an AI tool developed by 17-year-old Edward Kang. Using retinal photographs, RetinaMind was designed to distinguish individuals with autism, ADH, and those with neither diagnosis, reporting approximately 89% accuracy. The system also highlights retinal regions influencing its predictions, offering researchers opportunities to investigate potential biological differences. Although still experimental and not clinically validated, RetinaMind illustrates how this emerging technology could transform neurological screening (Waseem, 2026).

Perhaps the eyes really are windows to something deeper, offering a glimpse not only into how we see the world, but into the health of our brains.


References

Choi, H. et al. (2025) ‘Retinal fundus imaging as biomarker for ADHD using machine learning for screening and visual attention stratification’, npj Digital Medicine, 8, article 164. Available at: https://doi.org/10.1038/s41746-025-01547-9.

London, A., Benhar, I. and Schwartz, M. (2013) ‘The retina as a window to the brain—from eye research to CNS disorders’, Nature Reviews Neurology, 9(1), pp. 44–53. Available at: https://doi.org/10.1038/nrneurol.2012.227.

Sathianvichitr, K. et al. (2023) ‘Through the eyes into the brain, using artificial intelligence’, Annals of the Academy of Medicine, Singapore, 52(2), pp. 88–95. Available at: https://doi.org/10.47102/annals-acadmedsg.2022369.

Waseem, R. (2026) ‘This high schooler developed an A.I. tool to diagnose autism and ADHD using the retina’, Smithsonian Magazine, 11 May. Available at: https://www.smithsonianmag.com/innovation/this-high-schooler-developed-an-ai-tool-to-diagnose-autism-and-adhd-using-the-retina-180988694/

Share this article

Share to Facebook
Share to X
Share to LinkedIn

Written by