What AI can (and can’t) tell us about the next frontier in myopia detection and management
Can AI really see the future? From deep learning models to fundus “fortune-telling,” here’s how artificial intelligence is transforming myopia management into a smarter, sharper and more personalized science.
These days, technology seems to predict just about everything, from your next Netflix obsession to whether you’ll need an umbrella tomorrow. So it was only a matter of time before artificial intelligence (AI) set its sights on the future of our patients’ vision.
Think of it as a high-tech crystal ball: AI and predictive analytics are reshaping the way eye care professionals tackle myopia. What was once a reactive game of catch-ups is evolving into a smarter, more proactive and highly personalized approach to care.
Demystifying AI
For some optometrists, hearing the term “artificial intelligence” might trigger visions of robot overlords or mysterious black-box algorithms making clinical calls. But according to Dr. Li Lian Foo, a consultant at Singapore National Eye Centre, it’s time to set the record straight…and maybe lower your heart rate a little.
“One common misconception is that AI will eventually replace the role of optometrists,” Dr. Foo explained. “In reality, AI is best seen as an assistive tool that augments, rather than replaces, clinical judgment.”
That fear of being replaced isn’t the only misconception she encounters. “Another is that AI is a ‘black box’ that cannot be trusted or understood. Additionally, some may assume AI is only for high-tech institutions or research settings,” she added. “In truth, scalable AI tools are being integrated into everyday practice, including community and primary eye care, to improve screening, triage and access, especially in underserved areas.”
Still feeling a little wary? You’re not alone. But Dr. Foo offers a helpful reframe: Think of AI as your extra set of eyes. “It can rapidly analyze large volumes of data, highlight subtle changes that might otherwise be missed, or flag high-risk cases earlier,” she explained. “But it doesn’t have context, empathy or clinical intuition, which are qualities that only a trained eye care professional brings to patient care.”
And if all of this still sounds a bit sci-fi, Dr. Foo gets it. “It’s completely natural to feel a little uncertain when new technologies enter clinical practice, especially something as complex as AI,” she said. “Rather than taking over, AI serves as a supportive partner. It frees up our time from routine or repetitive tasks so we can focus more on what really matters: connecting with patients, making nuanced decisions and personalizing care.”
So, if you’re worried AI is coming for your job, relax. It’s not taking your place. It’s just here to make your job easier, sharper and maybe even a little more fun.
READ MORE: Myopia Care in The Modern Age
Early detection
One of AI’s superpowers in the realm of myopia control is spotting trouble before it even knocks.
“Researchers are now leveraging AI and imaging technologies—especially retinal and optical coherence tomography (OCT) images—to uncover subtle structural changes in the eye that may signal early myopia development,” said Dr. Foo. “AI algorithms can be trained on large datasets of fundus images to detect patterns like changes in optic disc shape, retinal vessel trajectory or early elongation markers that are difficult for the human eye to quantify consistently.”
Here’s where things get especially fascinating: A 2025 study introduced a fresh AI-powered concept called “fundus refraction offset” (FRO), and it just might change the way we think about early myopia detection.1
FRO is an AI-derived metric that measures the gap between what an algorithm thinks the refractive error should be (based on features in the fundus image) and what it actually is.1
To build this tool, researchers trained a deep learning model using over 31,000 fundus images from healthy eyes, teaching it to predict spherical equivalent refraction (SER). When the model was let loose on new, unseen images, it could spot eyes with a fundus that looked more myopic than their actual refractive error suggested.1
Even more intriguing? Eyes with a more negative FRO—meaning the fundus looked more myopic than expected—also showed signs like macular thinning and lower choroidal vascularity index. And this was true regardless of their measured SER or axial length. In other words, FRO may help reveal which eyes are structurally more at risk, even when their numbers look deceptively normal.1
“The fundus refraction offset might capture hidden structural changes in the eye that are not yet clinically apparent but are predictive of future myopia progression,” said Dr. Foo. “What is particularly exciting is that this metric could serve as an early warning signal. If the AI detects a fundus pattern that looks more myopic than the current refraction suggests, it could indicate a higher risk of progression in the near future.”
This means eye care professionals might soon be able to intervene before refractive error shifts into high gear, giving us a head start in the race against childhood myopia.
The promise and puzzles of genetic data
Imaging isn’t the only place AI is flexing its predictive muscles. Genetics is the next big frontier in the quest to outsmart myopia.
“Incorporating genetic data into AI models for myopia prediction is a very promising frontier,” said Dr. Foo. “We know that myopia is influenced by both genetic and environmental factors, but until recently, clinical practice has focused mostly on lifestyle and refractive trends.”
The idea is to combine polygenic risk scores with other data inputs to create a fuller, more personalized risk profile for each child. “This means we could one day identify children who are genetically predisposed to develop high myopia, even before they show clinical signs on refraction or imaging,” Dr. Foo explained.
But as with any fortune-telling device, the signals can sometimes get fuzzy. A 2025 review by Liu et al. pointed out that “existing studies have shown that the added value of genetic data in myopia prediction is very limited, and research results are inconsistent.” Some studies reported that genetic risk scores didn’t move the predictive needle much, while others found that they offered an edge over relying on family history alone.2
Still, researchers are far from giving up on the genetic angle. In fact, Ruamviboonsuk et al. recently spotlighted four potential biomarkers (NR1D1, PPP1R18, PG8D2 and PPP1R3D) using machine learning tools. These findings lay the groundwork for future clinical validation and may yet give genetic data a bigger role in shaping personalized prevention strategies.3
READ MORE: Outsmarting Myopia
AI’s data-driven predictions
AI isn’t just good at spotting early signs of myopia, it’s also becoming a bit of a good fortune teller when it comes to predicting how myopia will progress. A 2024 study, published in Precision Clinical Medicine, developed a multivariate linear regression algorithm that crunched data from more than 612,000 medical records across five independent cohorts.4
Its mission? Predict myopia progression and the risk of developing high myopia. Spoiler alert: it crushed it.
The model achieved an R² value of 0.964 in internal validation and continued to impress across external datasets. For predicting the risk of high myopia, the algorithm clocked in with an area under the curve (AUC) of 0.99—a nearly clairvoyant level of accuracy.4
So what’s behind the magic? Unlike traditional risk assessments that might look at just a few variables (age, current refraction, maybe parental myopia), AI models pull from a much deeper data well.
“By integrating data from multiple sources, such as a child’s age, rate of progression, axial length, lifestyle habits, genetic risk and even retinal imaging, AI can identify unique risk profiles and recommend tailored treatment strategies,” explained Dr. Foo. “AI can help us match the right intervention to the right patient, at the right time.”
The same study also revealed some key clues about who’s at greater risk of high myopia. Kids whose myopia worsened by more than 1.00 D per year hit the high myopia milestone in a median of 3.39 years. Slower progressors had a little more breathing room, about 4.05 years.4
Age of onset matters, too. Children who became myopic between the ages of three to seven reached high myopia in a median of 3.13 years, compared to 4.01 years for those with onset between eight and 18.4
The takeaway? The earlier and faster the progression, the higher the stakes…and the stronger the case for jumping in early with tailored, proactive care.
Personalized treatment
Among all the breakthroughs AI brings to myopia management, its ability to personalize treatment strategies might be the clearest signal yet that we’re stepping into the future.
Until now, many treatment plans have followed a fairly standardized path, largely based on a child’s age and current refraction. But with AI in the picture, we’re trading in generic protocols for something far more tailored.
“AI has the potential to revolutionize how we approach myopia by moving from a ‘one-size-fits-all’ model to truly personalized care,” said Dr. Foo. “It can help us match the right intervention to the right patient, at the right time. It can also predict who is likely to progress rapidly and may require more intensive follow-up, versus those who are stable and need less frequent visits.”
That level of insight means clinicians aren’t just making educated guesses, they’re guided by real-time, data-driven foresight. As Dr. Foo explained, “We are essentially moving toward a feedback loop where AI does not just guide the start of treatment but helps optimize it over time. By continuously analyzing longitudinal data, such as changes in axial length, refraction progression, compliance, side effects and lifestyle factors, AI can help identify which patients are responding well to a particular treatment and which are not.”
This kind of ongoing analysis lets clinicians fine-tune treatment decisions, rather than sticking to a rigid plan. “This kind of adaptive learning system is key to realizing precision medicine in myopia management, not just treating the condition, but treating the individual,” Dr. Foo noted.
Challenges and considerations
Of course, even the clearest crystal ball might come with a few cracks. As promising as AI is for myopia management, it is not without its blind spots.
“One major concern is bias. If AI models are trained on datasets that lack diversity in ethnicity, geography or socioeconomic background, the predictions may not generalize well across different populations,” she explained. “This could lead to unequal care or missed risks in underrepresented groups.”
Another red flag? Data privacy. As more personal information, like genetic data and high-resolution images, gets pulled into the mix, the stakes get higher.
“As we embrace AI in myopia management, we must remain vigilant about its limitations,” cautioned Dr. Foo. “Transparency, proper validation and ethical oversight are essential to building trustworthy AI systems that are safe, equitable and clinically meaningful.”
Embracing the crystal ball
Like any crystal ball from the pages of folklore, AI’s predictive powers in myopia management aren’t flawless. They come with their share of blind spots and a clear need for expert interpretation. But unlike their mythical counterparts, these tools are backed by data, clever algorithms and some serious scientific legwork.
Right now, eye care professionals are at a crossroads. We can approach AI with arms crossed, wary of its imperfections and potential pitfalls. Or, we can see it for what it really is: a powerful, data-driven ally in our fight against the global rise of myopia. Not a magic solution, but a high-tech crystal ball offering early clues and sharper insights into our patients’ visual futures.
READ MORE: WSPOS Myopia Consensus Statement 2025: What Parents and Practitioners Need to Know
The smartest way forward? A bit of both. Use AI for what it does best—analyzing, predicting, flagging risk—and pair it with what only humans can bring to the table: clinical judgment, compassion and ethical decision-making. Together, this combo could move myopia care from reactive firefighting to truly proactive, personalized prevention, saving millions from the long-term consequences of high myopia.
The crystal ball is here. The real question is: Are we ready to look into it?
Editor’s Note: A version of this article was first published on COOKIE Issue 20.
References
1. Yii F, MacCormick IJC, Strang N, et al. Fundus refraction offset as an individualized myopia biomarker. JAMA Ophthalmol. 2025:e251513. [Epub ahead of print.]
2. Liu N, Li L, Yu J. Application of artificial intelligence in myopia prevention and control. Pediatr Investig. 2025;9(2):114-124.
3. Ruamviboonsuk V, Lanca C, Grzybowski A. Biomarkers: Promising tools towards diagnosis, prognosis, and treatment of myopia. J Clin Med. 2024;13(22):6754.
4. Li J, Zeng S, Li Z, et al. Accurate prediction of myopia progression and high myopia by machine learning. Precis Clin Med. 2024;7(1):pbae005.
