Cataracts are still a leading cause of blindness and, with an ageing population, that burden will only grow [1]. Increasingly experts are leaning on technology such as ‘Surgery 4.0’ [2] – where smart machines and artificial intelligence (AI) slide into the surgical workflow – this is no longer sci-fi; it’s happening now [3]. Digital biometry, image-guided diagnostics and routine surgical video capture create the perfect dataset for AI to chew on.
New systems help with everything from preoperative risk prediction (and therefore, stratification) to intraoperative workflow analysis, surgical training and postoperative outcome optimisation. The goal isn’t to replace the surgeon; it’s to reduce variability, back up decision-making and enable more personalised care. Here, AI means a range of techniques: classic machine learning, deep learning, convolutional neural networks – all analysing complex ophthalmic data to improve precision and streamline workflows [4].
Refractive precision: AI in IOL power calculations
Getting patients close to emmetropia or their chosen refractive outcome is the main aim of modern cataract surgery. Traditional formulas like SRK/T or Holladay 1 are useful, but they’re built on static vergence models and can stumble in outlier eyes, like very long or very short axial lengths.
From static formulas to dynamic learning
AI changes the game. Models such as random forests and gradient boosting machines aren’t limited to two or three inputs. They can fold in a dozen or more variables – lens thickness, white-to-white distance and so on – to better predict effective lens position (ELP) [1]. Even more importantly, they pick up on nonlinear anatomical relationships that simple equations miss.
Comparative performance
The evidence is encouraging. Recent 2024–2025 studies show AI-based formulae deliver a higher proportion of eyes within ±0.5D of target. For instance, ensemble AI methods like Kane reported a mean absolute error (MAE) of about 0.272D with roughly 88.2% of eyes within 0.5D. Pattern-recognition AI (Hill-RBF 3.0) showed an MAE near 0.284D and 87.5% within 0.5D. Hybrid approaches such as PEARL-DGS came in at 0.290D and 85.1%. By contrast, well-established traditional formulas like Barrett Universal II had an MAE of about 0.312D (81.0% within 0.5D), and SRK/T lagged at around 0.405D (71.3% within 0.5D) [1,5].
AI in post-refractive cataract surgery
One of the most exciting and challenging applications is in eyes that had previous corneal refractive procedures. Those altered corneae wreck the assumptions most formulas rely on: curvature and anterior-posterior relationships change, higher-order aberrations creep in and the usual tricks can lead to refractive surprises. AI models can integrate multimodal data – past refractive history, tomography, wavefront measures, modern biometry – to better estimate ELP and postoperative refraction. Instead of a one-size-fits-all adjustment, machine learning can spot patterns across lots of post-refractive eyes. That’s huge when patients expect spectacle independence, especially with premium lenses on board.
AI in premium cataract practice
It takes careful evaluation when planning toric, multifocal or extended-depth-of-focus intraocular lenses (IOLs) [6], because the margin for error is tiny, but AI can help. By analysing corneal metrics, posterior corneal astigmatism, axis stability and ocular-surface factors, AI tools can optimise toric planning and lower residual astigmatism. They can also suggest which eyes are likely to tolerate multifocal optics versus those that’d do better with a more conservative lens choice, integrating biometric data, lifestyle factors and risk signals for dysphotopsia or contrast loss. Wavefront and tomography-guided decision support, for example, can flag when higher-order aberrations make premium optics a poor idea. Put bluntly: AI gives you an extra analytical layer to refine decisions where patient expectations are highest – again, assisting the surgeon, not overruling them.
Training the resident
Cataract surgery has a steep learning curve, and AI is making training more objective and targeted.
Objective performance scoring
Traditional skill assessment is subjective and inconsistent. Enter automated surgical analysis, sometimes called ‘PhacoTracking’, which measures instrument path length, number of movements, total case time and other metrics. AI-produced scorecards point out who’s wasting motion or rushing steps. For instance, motion tracking can flag a resident who moves forceps 50% more than an experienced consultant for the same task [1]. Kinematic analysis that studies velocity and economy of motion can separate novices from masters with about 96% accuracy.
"The most meaningful advances may arise not from autonomous surgery but from intelligent assistance"
High-fidelity simulation
Virtual reality simulators such as EyeSi now use AI alongside haptics to teach technique. The system senses stress on zonules and gives immediate feedback – physical and visual – so trainees learn the feel of tissue without putting a patient at risk [7]. That kind of repetitive, safe practice shortens the learning curve.
Operational excellence: AI workflow systems
The bottleneck in many services isn’t the surgery itself but admin and logistics, and AI operating systems aim to unclog that.
Intelligent surgical planning
Cloud platforms (Alcon Smart Cataract, Zeiss Veracity, Bausch + Lomb Eyetelligence) link diagnostics to the cloud, automating data entry. That reduces transcription errors – which historically have caused wrong-power IOL implants – and helps flag high-risk cases early so schedules and staffing can be adjusted [8]. In short: better data flow, fewer avoidable mistakes.
Intraoperative adaptive systems
On the intraoperative front, platforms like ALLY (LENSAR) adapt laser fragmentation in real time to cataract density, optimising energy delivery and lowering the risk of postoperative corneal oedema [9]. These are small but meaningful efficiencies that translate into better outcomes.
Postoperative monitoring and complications
AI shifts care from reactive to predictive. Models can detect posterior capsular opacification from imaging with very high accuracy area under the curve up to 0.97 in some reports. Natural language processing assistants can do symptom screening and triage; in 2024 trials, these systems showed excellent sensitivity for spotting issues requiring urgent care [1]. Remote follow-up becomes more feasible, and low-risk patients can avoid unnecessary clinic visits. There is also the remote patient-monitoring AI software that was facilitated by the Dora (Ufonia) system, utilising natural language processing to evaluate post-cataract surgery outcomes and identify potential complications such as endophthalmitis or persistent pain. Through structured telephonic assessments, Dora can escalate concerning symptoms to the clinical team in real-time [10].
What AI still can’t do in cataract surgery
Let’s be clear: AI is powerful, but it’s not a human surgeon. It can crunch data and spot patterns, but it can’t shoulder the responsibility of surgical judgement. It struggles with uncertainty, can’t improvise intuitively during unexpected intraoperative events, and doesn’t sense the subtle tactile or behavioural cues that often guide split-second decisions. Most of all, AI can’t counsel a patient, manage expectations or be legally and ethically accountable when things go wrong. The surgeon remains the decision-maker. Always.
Conclusion
AI is moving from novelty to necessity in cataract care. The most meaningful advances may arise not from autonomous surgery but from intelligent assistance: improved prediction of outcomes, enhanced surgical training, real-time workflow awareness and data-driven quality improvement. That said, successful adoption depends on clean data, transparent algorithms, ethical oversight and thoughtful integration into clinical teams. Done right, AI will amplify what surgeons already do well: make care safer, more precise and more personalised. Done poorly, it risks adding complexity without benefit. So, we proceed, cautiously optimistic, and eager to see how these tools evolve.
References
1. Ahuja AS, Paredes III AA, Eisel MLS, et al. Applications of artificial intelligence in cataract surgery: a review. Clin Ophthalmol 2024;18:2969–75.
2. Feussner H, Ostler D, Kranzfelder M, et al. Surgery 4.0. In: Thuemmler C, Bai C (Eds.) Health 4.0: How Virtualization and Big Data are Revolutionizing Healthcare. Springer, Cham. 2017:91–107.
3. Yuan A, Lee CS. Real-time augmented reality: the next frontier for ophthalmic surgery. JAMA Ophthalmol 2022;140(2):177–8.
4. Olawade DB, Weerasinghe K, Mathugamage MDDE, et al. Enhancing ophthalmic diagnosis and treatment with artificial intelligence. Medicina (Kaunas) 2025;61(3):433.
5. Stopyra W, Voytsekhivskyy O, Grzybowski A. Prediction of seven artificial intelligence-based intraocular lens power calculation formulas in medium-long Caucasian eyes. Life (Basel) 2025;15(1):45.
6. Chen JL, Al Mohtaseb ZN, Chen AJ. Criteria for premium intraocular lens patient selection. Curr Opin Ophthalmol 2024;35(5):353–8.
7. Dormegny L, Lansingh VC, Leejay A, et al. Virtual reality simulation and real-life training programs for cataract surgery: a scoping review of the literature. BMC Med Educ 2024;24(1):1245.
8. Lobanoff M. AI-powered cataract surgical planning. CRS Today 2026 [Online]:
https://crstoday.com/articles/jan-2026/
ai-powered-cataract-surgical-planning
9. Weinstock R. AI-related innovations for laser cataract technology mark a milestone in history of device. Ophthalmology Times 2024 [Online]:
www.ophthalmologytimes.com/view/
ai-related-innovations-for-laser-cataract-technology
-mark-a-milestone-in-history-of-device
10. Khavandi S, Lim E, Higham A, et al. User-acceptability of an automated telephone call for post-operative follow-up after uncomplicated cataract surgery. Eye (Lond) 2023;37(10):2069–76.
[All links last accessed April 2026]
Declaration of competing interests: None declared.
