Getting a Hold of AI

Ready or not, an AI boom is upon us, affecting nearly every aspect of daily life. Millions of people are now using AI-based apps trained on billions of data points, generating trillions of dollars for investors. Meanwhile, AI is moving rapidly from research promise to clinical reality in ophthalmology.

With AI applications already entering clinical practice and many more in development, every ophthalmologist should have a working understanding of the technology. As a general summary, artificial intelligence is an umbrella term for software systems that can perform tasks normally requiring human intelligence. Machine learning allows systems to identify patterns from large data sets. Large language models analyse and generate language. Generative AI creates new content, including text, images, video, and code. Multimodal AI can combine images, text, and other forms of data. Together, these technologies are creating a growing range of clinical, administrative, educational, and research applications.

As a specialty built on image analysis, structured data, and precision measurement, ophthalmology is well suited to taking advantage of AI. Indeed, ophthalmology was an early adopter. In 2018, IDx-DR (now Luminetics Core) became the first autonomous AI diagnostic system in medicine approved by the US Food and Drug Administration (FDA). That AI system screens large image libraries to identify patients at risk for diabetic retinal disease progression requiring referral.

Retina remains one of the most active areas for ophthalmic AI. Applications are being developed for age-related macular degeneration, diabetic macular oedema, retinal vein occlusion, geographic atrophy, inherited retinal disease, and treatment-response prediction.

Beyond retina

AI has since expanded into all areas of ophthalmology, including cataract and refractive surgery, cornea, and glaucoma. It is being used for everything from advanced research to smoothing patient flow in the office setting.

The most noticeable appearance of AI in the cataract and refractive field is in IOL power calculation. AI-based formulas including Hill-RBF, PEARL-DGS, Kane, and Zeiss-AI use machine learning to produce guidance on IOL selection. The ESCRS IOL Calculator (https://iolcalculator.escrs.org/) includes AI-based systems among its options.

A study by Woong-Joo Whang MD and colleagues indicates that AI could not only help improve IOL power calculation but also solve one of the most stubborn issues in cataract surgery—predicting an effective lens position (ELP). Their study suggests that using attention-enhanced deep learning models can optimise ELP prediction. 

Another exciting development in which AI plays a big role is robotic-assisted cataract surgery. In late 2025, Uday Devgan MD performed the first robot surgeries in humans using the Horizon Polaris system (in which he has declared a financial interest). That system uses microrobotic arms, 3D visualisation, and real-time machine learning to assist in routine cataract surgery.

“The next step for robotics is, can the robot give me another benefit? Can it help me visualise or see things I couldn’t otherwise see? Can the robot actually make me a safer surgeon? Like, no matter how dumb I am, don’t let me make this mistake. We can tell the system, don’t let me bring my instrument within 5 or 10 microns of the posterior capsule, ever,” Dr Devgan told EuroTimes.

The near future is not a robot replacing the cataract surgeon. Rather, it is a digitally guided operating room in which imaging, planning software, machine learning, and robotic motion control make surgery more predictable, measurable, and (potentially) scalable, he noted.

“I anticipate that before we realise it, sooner rather than later, we’ll have the robot with the ability to do an entire cataract case,” he predicted.

Cornea

The cornea field is also integrating AI into many aspects of practice. Its impact can be seen in keratoconus screening in particular. Convolutional neural networks trained on Scheimpflug tomography, topography, and other corneal imaging data are achieving high accuracy in detecting and grading keratoconus. Similar approaches are being explored for ectasia risk prediction, refractive surgery screening, and infectious keratitis triage.

“Artificial intelligence is very important for teaching, research, and clinical use,” Andreia Rosa MD, PhD explained. “We are using it for keratoconus screening and establishing probability of progression and for corneal infections progression detection. Many tools we use already have AI incorporated.”

Glaucoma

Glaucoma is another area where AI may be particularly useful because diagnosis and monitoring depend on multiple data streams: optic disc photographs, OCT retinal nerve fibre layer and ganglion cell analysis, visual fields, IOP, pachymetry, and more. AI systems are being studied for glaucoma detection, visual field interpretation, progression analysis, and risk stratification.

Everyday clinical practice

For most surgeons, one of the least favourite parts of clinical practice is the endless paperwork and documentation. AI can take on these demands with digital scribes, patient-facing chatbots, screening tools, and office management platforms.

“As ophthalmologists, we spend a significant proportion of our day on tasks that don’t require our clinical expertise: reviewing records, documenting consultations, navigating multiple imaging platforms, analysing data, and managing increasingly complex pathways,” Artemis Matsou MD said. “If AI can take on some of that work safely, it allows us to spend more time doing what only clinicians can do: examining patients, operating, communicating, and making complex decisions.”

She is interested in using AI to improve clinical pathways. In cataract surgery, for example, her team has introduced AI tools that support virtual triage, identify patients who are suitable for streamlined pathways, and reduce unnecessary hospital visits. AI-powered follow-up systems allow routine postoperative patients to be monitored safely while ensuring those with potential problems are escalated for clinician review.

Better regulate than never

The rapid expansion of AI in medicine also brings significant risks, including a lack of accountability, hidden bias, patient safety concerns, privacy risks, cybersecurity vulnerabilities, plagiarism, overreliance on automated outputs, and uncertainty about who is legally responsible when something goes wrong. The changes wrought by AI have been so rapid and dramatic that governments and professional groups have been slow to respond. Yet most agree medical AI needs guidelines, even guardrails.

To that end, the EU AI Act, passed in August 2024, recognises the risks posed by AI and proposes transparency and clinical accountability for its use. Some AI tools are likely to fall under the purview of European Medical Device Regulation (MDR) rules on production, clinical testing, and distribution of medical devices in the EU. In the UK, AI devices are required to go through the existing regulatory system, but an AI-specific system is in development.

“I am optimistic about AI in ophthalmology, but more cautious than excited. The specialty is clearly well suited to it because we rely so heavily on imaging and longitudinal data. But I don’t think the best version of this future is one where machines replace ophthalmologists,” commented Sian Liu MD, PhD. “The real value will probably be in less dramatic tools that reduce administrative tasks, help detect disease earlier, support triage, and give clinicians better information while keeping the doctor–patient relationship intact.”

Uday Devgan MD is founder of Devgan Eye Surgery, Los Angeles, US, and creator of cataractcoach.com. devgan@gmail.com

Woong-Joo Whang MD is Assistant Professor of Ophthalmology, Catholic University of Korea, South Korea. olokl@nate.com

Artemis Matsou MD, MRCP (UK), FEBO is Consultant Ophthalmologist and Cataract Lead at Queen Victoria Hospital, East Grinstead, UK. art.matsou@gmail.com

Andreia Rosa MD, PhD is a specialist in the Cornea and Refractive Surgery Section of the University Hospital of Coimbra, Portugal.

Sian Liu MD, PhD is NIHR Academic Clinical Lecturer, UCL Institute of Ophthalmology, London, UK. siyin.liu@ucl.ac.uk