AI is Reshaping Preventive Eye Care and Protecting Vision

Eyes, a key sensory organ, play an important role in every stage of our lives.

They feed information to the brain about the outside world. Without vision, we would struggle to perform daily tasks like reading, learning, walking, and interacting with our environment.

Good vision is essential for enjoying independence and, of course, the joys of life.

However, more than 2.2 billion people are unable to do so due to having a vision impairment, which happens due to an eye condition affecting the visual system and its functions. 

Cataracts, glaucoma, refractive errors, diabetic retinopathy, age-related macular degeneration (AMD), and presbyopia are some of the leading causes of vision impairment.

Vision impairment has grave consequences for one’s life, many of which can be mitigated through timely access to quality eye care. Beyond its impact on individuals, vision impairment also poses a significant financial burden, with the annual global cost of lost productivity estimated at $411 billion.

As a result, eye conditions that can cause vision impairment and blindness are the primary focus of eye care strategies.

AI in Ophthalmology

As vision impairment reduces a person’s quality of life and creates a considerable global economic burden, doctors, scientists, and researchers are turning to artificial intelligence (AI) to transform eye care.

The rapidly evolving technology has been improving business efficiency and data analysis across finance, manufacturing, retail, media, and healthcare.

In healthcare and medicine, the influence of AI has risen drastically in recent years. 

AI for Early Detection: DR, Glaucoma, and AMD

In ophthalmology, specifically, AI is enabling the early detection of diseases like glaucoma, diabetic retinopathy, and AMD through the analysis of retinal images and patient data. Here, AI tools enable personalized interventions and automated diagnosis and screening. 

The tech also provides surgical assistance, helping screen candidates, optimize techniques, minimize complications, and predict postoperative outcomes.

AI, according to experts, will play an increasing role in clinical and surgical practice in the future of anterior segment surgery, which focuses on the eye’s front structures to treat injuries and conditions like cataracts and corneal disorders.

Already, the technology is being used here to screen and diagnose cataract, classify surgical phases, and predict surgical timing to optimize OR workflow. Moreover, AI is being applied to the optimal calculation of IOL (intraocular lens) power, which refers to the refractive strength of the synthetic lens implanted during cataract surgery to replace the natural lens.

There is, after all, a lot of literature as well as imaging data collected over the years as a routine practice.

Swipe to scroll →

How LLMs Perform vs. Ophthalmologists

A study actually found that AI is surpassing non-specialist doctors’ ability¹ to assess eye problems.

Led by the University of Cambridge, the study reported that the clinical knowledge and reasoning skills of the popular large language model (LLM), GPT-4, are now approaching the level of specialist eye doctors.

GPT-4 was tested against expert eye doctors, unspecialized junior doctors, and trainees, with each one presented with as many as 87 scenarios concerning specific eye problems. The questions covered a wide range of eye problems, taken from a textbook that is used to test trainees but isn’t freely available on the internet, which makes it unlikely that GPT-4’s training datasets included that content.

The doctors had to choose a diagnosis or treatment advice from four options. As per the study, the AI model scored higher than junior doctors in the test and had about the same scores as trainees and experts, though the top-performing doctors scored higher than GPT-4. 

LLMs won’t be replacing healthcare professionals, though, researchers said, but noted that they can improve healthcare by providing diagnosis, advice, and management suggestions in certain contexts.

From Lab to Clinic: Real-World Screening Results

A recent study evaluated AI’s implementation in glaucoma detection² in real-world settings.

For this, they developed an automated retinal photography and AI-based screening system to assess its acceptability, feasibility, and accuracy. The study recruited people aged 50 years or above with their retinal images taken with an automated fundus camera and analyzed by AI. 

The AI system achieved an AUROC of 0.80, which shows the tech’s strong capability to distinguish between conditions. Sensitivity was 65%, representing actual cases correctly identified by the AI, while specificity was 94.6%, reflecting accurate classification of healthy individuals. Among 161 patients who were previously undiagnosed, 18 (11.2%) were identified as referable glaucoma. The study stated:

“Despite challenges such as lower sensitivity and image acquisition limitations, the system shows promise for opportunistic screening in primary care settings.” 

A review³ by the Department of Ophthalmology at Capital Medical University, China, meanwhile, explored the applications and challenges of AI in myopia. 

Currently, this condition affects over two billion people globally, and by 2050, almost half of the world’s population is expected to be affected by it. When left uncorrected, myopia can impair vision, disrupt education, and affect employment opportunities, while high myopia can cause permanent vision loss. So, early diagnosis is important in managing its progression and preventing any long-term visual damage.

Here, AI can offer a promising tool by analyzing complex medical data.

To detect myopia, AI models can be trained on large quantities of fundus photos and OCT images and then be taught to find changes in the retina that are associated with myopia. AI can also be trained to detect any behavioral changes associated with the onset of myopia. Even self-monitoring equipment like SVOne can use AI to detect refractive errors in the eyes. In order to discover risk factors, techniques like logistic regression, support vector machine, and XGBoost can be used. 

While AI can help clinical practice and policymaking control myopia, it isn’t without challenges.

“By building high-quality datasets, improving the model’s capacity to process multimodal image data, and improving human-computer interaction capability, the AI models can be further improved for widespread clinical application.”

– Dr. Jifeng Yu

Addressing the Disparity

While visual problems are widespread, the impairment is more prevalent in low- and middle-income countries (LMICs) compared to high-income regions. By allowing more people to be screened, AI can help bridge the gap in regions with limited access to specialized eye care.

For this, tech giant Google has created an AI model called ARDA (Automated Retinal Disease Assessment), and recently, it licensed the AI model for detecting diabetic retinopathy to health care technology companies in Thailand and India.

“They will be setting up their own business models, but on the side, they will also be delivering screenings to people who need it the most but can’t afford it. Blindness from diabetic retinopathy is completely preventable, and the fact that we have not been able to do effective screening in some of these places shouldn’t be forgiven.”

– Sunny Virmani, project manager at Google Health

Diabetes, which is becoming more common in LMICs, can severely impact vision by causing blurry vision, diabetic macular edema, glaucoma, and diabetic retinopathy. The latter results from excessive blood sugar damaging blood vessels, leading to fluid leakage into the eye.

Diabetic retinopathy can cause changes in vision, and eventually, a person may become blind. However, early diagnosis and treatment reduce the risk by as much as 98%. But only a small number of people with diabetes get screened.

So, over a decade ago, Dale Webster, director of research at Google Health, along with his colleagues, began to test AI’s ability to diagnose disease from medical images. 

This resulted in ARDA, which can diagnose the disease as effectively as an ophthalmologist.

For the AI model, the Google team screened 7,651 people in three regions in Thailand between 2018 and 2020, with ARDA achieving an accuracy of 94.7%, showing “that these tools are safe and effective.”

Click here for a list of the top ten rare disease specialist stocks.

AI’s Breakthrough in Managing Keratoconus

Amidst all this progress being made, researchers have now developed an AI that can successfully predict which keratoconus patients are likely to go blind, thus requiring early treatment and monitoring, years before doctors can. By doing so, the technology can reduce unnecessary procedures and prevent vision loss.

Keratoconus is a progressive eye condition with no known cause.

In this condition, the cornea gets contorted. It is the clear, dome-shaped layer covering the iris and pupil that allows light to enter and helps to focus it for clear vision.

So, when the cornea thins out and bulges into a cone shape, that’s called keratoconus. The change in the cornea’s shape puts the rays of light out of focus, causing distorted vision. Other symptoms include glare, light sensitivity, and blurry vision. This makes daily tasks like driving or reading difficult.

This disease often develops in the late teens or early 20s and progresses over time.

While the exact cause of the disease is unknown, it may be genetic, with 1 out of 10 people with keratoconus having a parent who also has it. Keratoconus is also related to excessive eye rubbing, eye allergies, corneal thinning due to collagen loss, and connective tissue disorders.

Typically affecting both eyes, the disease can actually lead to very different vision and symptoms between the two eyes.

The symptoms of keratoconus slowly get worse over a period of ten to twenty years. In the early stage, the symptoms may include eye redness or swelling, increased sensitivity to light and glare, mild blurring of vision, and slightly distorted vision. In later stages, symptoms often include increased nearsightedness or astigmatism and more blurry and distorted vision.

During the early stages, vision problems can often be corrected with glasses or contact lenses, but in later stages, rigid gas-permeable contact lenses may be required.

But if not treated on time and the condition becomes more severe, cornea transplants, Intacs (small corneal implants), and corneal cross-linking (CXL) may be needed. Now, to diagnose keratoconus, doctors monitor patients over time.

During the routine eye exams, an ophthalmologist will examine the cornea and may use specialized imaging to measure the curvature, which will show any changes in its shape.

“Keratoconus is a manageable condition, but knowing who to treat, and when and how to give treatment is challenging. Unfortunately, this problem can lead to delays, with many patients experiencing vision loss and requiring invasive implant or transplant surgery.”

– Dr. José Luis Güell, ESCRS Trustee and Head of the Cornea, Cataract and Refractive Surgery Department at the Instituto de Microcirugía Ocular, Spain

But researchers have now achieved a breakthrough that could reshape eye care by enabling AI to predict which keratoconus patients need urgent corneal treatment before irreversible damage sets in, potentially saving eyesight and reducing transplants.

The study was recently presented at the 43rd Congress of the European Society of Cataract and Refractive Surgeons (ESCRS).

Saving Human Vision & Healthcare Resources

Conducted by Dr. Shafi Balal and colleagues at Moorfields Eye Hospital NHS Foundation Trust and University College London (UCL), the study used AI to analyze images of patients’ eyes and combined them with other data to predict which keratoconus patients required immediate treatment and which could continue with monitoring.

“Keratoconus causes visual impairment in young, working-age patients, and it is the most common reason for corneal transplantation in the Western world.”

– Dr. Balal

With just one single treatment called ‘cross-linking’, the progression of the disease can be halted. The therapy involves using ultraviolet light and vitamin B2 (riboflavin) drops to stiffen the cornea.

The cross-linking treatment, however, needs to be done before scarring becomes permanent, which removes the need for a cornea transplant. It is actually successful in over 95% of cases. According to Dr. Balal:

“However, doctors cannot currently predict which patients will progress and require treatment, and which will remain stable with monitoring alone. This means patients need frequent monitoring over many years, with cross-linking typically performed after progression has already occurred.” 

Hence, the AI can diagnose keratoconus on time.

For their AI, the study utilized a group of patients referred to Moorfields Eye Hospital for keratoconus assessment and monitoring, including scanning the eye front with OCT to examine its shape.

Optical coherence tomography (OCT) is a non-invasive imaging method that uses light waves to take high-resolution, cross-sectional pictures of the retina. The technique is widely used in ophthalmology to diagnose conditions like macular degeneration, glaucoma, and diabetic retinopathy.  

The researchers studied 36,673 OCT images of 6,684 patients using AI. The researchers found that their AI model can accurately predict whether a patient’s condition will remain stable or deteriorate, based solely on the patient’s first visit.

What this means is that from the initial routine consultation, AI can help doctors predict which patients are likely to experience progression, allowing for early treatment before progression and secondary changes occur.

Using the AI model, the researchers sorted patients into two groups. One group was low-risk, comprising two-thirds of patients who did not require treatment. The other was the high-risk group, with one-third of patients who needed immediate cross-linking treatment.

Upon including the data from the second hospital visit, the AI model could classify up to 90% of patients.

This makes the study the first of its kind to get such a high level of accuracy in predicting the risk of keratoconus progression using a combination of scans and patient data, said Dr. Balal. He added:

“Our research shows that we can use AI to predict which patients need treatment and which can continue with monitoring.” 

The study, Dr. Balal noted, involved a large group of patients who were monitored for two years or more. The results suggest that patients with high-risk keratoconus could receive preventative treatment before their condition progresses, helping to prevent vision loss and reduce the need for complicated corneal transplant surgery and its recovery burden.”

Meanwhile, low-risk patients can avoid unnecessary monitoring, which will save healthcare resources. 

“The effective sorting of patients by the algorithm will allow specialists to be redirected to areas with the greatest need.”

– Dr. Balal

The researchers are currently working on developing an even more powerful AI algorithm, which will be trained on millions of eye scans. The algorithm can be tailored for specific tasks, such as predicting keratoconus progression, as well as detecting inherited eye diseases and eye infections.

If the AI algorithm “consistently demonstrates its effectiveness, this technology would ultimately prevent vision loss and more difficult management strategies in young, working-age patients,” said Dr. Güell, who was not involved in the research.

The algorithm will now be going under further safety testing before it can be deployed in the clinical setting.

Investing in AI-driven Eye Care

Alcon AG (ALC -1.81%) is a Switzerland-based eye care company that researches, manufactures, and sells a suite of eye care products through Surgical and Vision Care for conditions like cataracts, glaucoma, retinal diseases, and refractive errors.

Alcon AG (ALC -1.81%)

With a market cap of $39.6 billion, ALC shares are currently trading at $77.78, down 8.81% YTD and about 23% from its peak above $100 late last year. With that, it has an EPS (TTM) of 2.25 and a P/E (TTM) of 34.41, while the dividend yield offered is 0.43%.

For Q2 2025, Alcon reported a 4% increase in sales to $2.6 billion. Diluted EPS for the quarter was $0.35.

The company generated $889 million in cash from operations during this period, while $681 million in free cash flow was posted in the first half of 2025. Alcon returned $287 million to shareholders.

Noting “robust early demand” for new products like Tryptyr, Systane Pro PF, Precision7, PanOptix Pro, Voyager, and Unity VCS, CEO David J. Endicott said:

“While it’s still early, these launches position us to accelerate top-line growth, generate cash, and deliver long-term value for our shareholders.”

Click here to learn all about investing in artificial intelligence.

Latest Alcon AG (ALC) Stock News and Developments

Conclusion

With tens of millions of people around the world at risk of vision impairment or blindness, AI’s ability to predict disease progression and guide treatment decisions marks a new era of preventive eye care. As algorithms become more refined, AI has the potential to empower clinicians to help protect vision, reduce healthcare costs, and enhance the quality of life.

Click here for a list of new advances targeting vision loss.

References:

<sup>1. Thirunavukarasu, A. J., Mahmood, S., Malem, A., Foster, W. P., Sanghera, R., Hassan, R., Zhou, S., Wong, S. W., Wong, Y. L., Chong, Y. J., Shakeel, A., Chang, Y.-H., Tan, B. K. J., Jain, N., Tan, T. F., Rauz, S., Ting, D. S. W., & Ting, D. S. J. (2024). Large language models approach expert-level clinical knowledge and reasoning in ophthalmology: A head-to-head cross-sectional study. PLOS Digital Health, (Version of Record), published 17 April 2024. https://doi.org/10.1371/journal.pdig.0000341
2. Jan, C. L., Joseph, S., Vingrys, A. J., et al. (2025). Prospective pragmatic trial of automated retinal photography and AI glaucoma screening in Australian primary care. npj Digital Medicine, 8, 386. (Version of Record), published 1 July 2025. Received 9 March 2025; accepted 2 June 2025. https://doi.org/10.1038/s41746-025-01768-y
3. Liu, N., Li, L., & Yu, J. (2025). Application of artificial intelligence in myopia prevention and control. Pediatric Investigation, (Version of Record), published 18 March 2025. https://doi.org/10.1002/ped4.70001</sup>