When artificial intelligence in healthcare is discussed in public, the conversation often revolves around documentation. AI assistants that convert speech to text save healthcare professionals time, and that is important. However, much less attention has been paid to AI that supports the diagnostic process itself. Eye health is a good example of this.
Tools for eye diseases already exist
AI solutions that support diagnostics are regulated medical devices whose safety and performance must be demonstrated before they can enter the market. In ophthalmology, several such tools are already available. In the United States, AI has been permitted to independently detect diabetic retinopathy since 2018.
In Europe, CE-marked AI tools can identify not only diabetic retinopathy but also changes associated with glaucoma and age-related macular degeneration from retinal images. Some of these tools support clinicians’ decision-making, while others operate independently during the screening phase.
New screening pathways
Finland has approximately half a million people with diabetes whose retinas are photographed regularly, typically every two to four years. The Finnish Current Care Guidelines for diabetic retinopathy note that AI is likely to be used to automate screening in the coming years, as the number of people requiring screening is growing faster than available resources.
AI could serve as the first reviewer of retinal images: normal findings could be cleared quickly, allowing ophthalmologists to focus on patients who need further attention. The same model could be applied to monitoring people at risk of glaucoma, such as those with elevated intraocular pressure. Optical stores already have the necessary equipment and expertise, making them a natural and cost-effective setting for imaging.
The eye is a window to the whole body
The eye is the only place in the human body where blood vessels and nerve tissue can be imaged non-invasively with micrometer-level precision. The field known as oculomics takes advantage of this. The same retinal image can reveal signs of cardiovascular disease, kidney disease, and neurodegenerative disorders.
In Europe, CE-marked algorithms are already available to assess cardiovascular risk from retinal images. In September 2026, researchers at Moorfields Eye Hospital reported that thinning visible in retinal layer scans was associated with the later development of atrial fibrillation. However, evidence in oculomics remains largely observational, meaning that such findings indicate risk rather than provide a diagnosis.
Technology alone is not enough
AI will not replace healthcare professionals; it will support their work. The benefits will only be realized when care pathways are clearly defined, data flows seamlessly between systems, and the expertise of optometrists is fully utilized. Vision impairment already costs healthcare systems around €2 billion annually, making early detection and prevention not only a health issue but also an economic one.
Documentation AI saves time. Diagnostic AI can save sight, or even lives.
Juha Jokinen
Director, Sales & Customer Success, Topcon Healthcare
The author works for Topcon Healthcare, an ophthalmic equipment manufacturer. The views expressed are the author’s own.
Resources
- Huemer J ym. Atrial fibrillation and retinal imaging-based oculomics. PLOS Digital Health 2026;5(9):e0001661. https://doi.org/10.1371/journal.pdig.0001661
- The Current Care Guideline on diabetic retinopathy (updated 11.6.2024). https://www.kaypahoito.fi/kht00052
- Hautala N. Päivitetty diabeettisen retinopatian Käypä hoito -suositus – mitä uutta? Diabetes-lehti 14.2.2025.
