Envision Summit 2026: Clinical sessions highlight AI's expanding role across ophthalmology

At Envision Summit 2026 in Rio Grande, Puerto Rico February 12-16, leaders in cornea, retina, cataract surgery, and clinical research convened to examine how artificial intelligence (AI) is reshaping ophthalmic care, from refractive screening and intraocular lens (IOL) calculations to diabetic retinopathy (DR) screening and AI-driven clinical trials. Across sessions, speakers emphasized that AI is not a replacement for physician judgment, but a tool to enhance precision, efficiency, and scalability.

Diabetic retinopathy screening: Implementation realities

In the first session, “Ophthalmic AI and Innovation,” Roomasa Channa, MD, FASRS, of the University of Wisconsin, and Shameema Sikder, MD, of Wilmer Eye Institute presented a comprehensive overview of artificial intelligence (AI) in ophthalmology across 2 domains: diabetic retinopathy (DR) screening and cataract surgery. In DR screening, it outlines the full lifecycle of AI development from identifying the clinical need amid the growing global diabetes burden and workforce shortages, to algorithm development, validation, FDA approval, and real-world implementation. While prospective trials demonstrated high sensitivity and specificity for detecting referable DR and diabetic macular edema, real-world data reveal implementation challenges, including geographic concentration in higher-income areas, high rates of non-diagnostic images, workflow barriers, financial constraints in federally qualified health centers, and variable follow-up adherence. Modeling data suggest that high AI screening uptake, particularly when paired with strong follow-up compliance, could substantially reduce vision loss at the population level, though disparities remain a concern.

Sikder discussed AI in cataract surgery, exploring surgical autonomy, computer vision–based instrument and phase recognition, automated annotation tools, and real-time analytics to support intraoperative decision-making. It emphasizes that while fully autonomous surgery remains distant, near-term applications include AI-guided assistance, augmented reality overlays, robotic integration, and decision support systems. Across both use cases, the overarching message is that AI’s value extends beyond diagnostic accuracy to thoughtful implementation, equity, safety, workflow integration, and measurable improvements in patient-centered outcomes.

“I think 1 of the keys is being able to create data sets that represent different institutions, so that you're not learning from one specific camera system, one specific hospital system, and then creating a solution that applies only to that specific entity. ultimately, one of the greatest goals that we would have in the work that we're doing is dissemination … having technology that can make it accessible to all is critical,” said Sikder.¹

Complementing this perspective, Laxmi V. Devisetty, MD (St. Luke's University Health Network) stateded that AI in retina “outputs probabilities, not diagnoses” and “will not replace retinal expertise or clinical judgment.” In the second session “Incorporating AI Into Your Clinical Practice,” Devisetty opened with reviewing how AI is currently being integrated into DR screening, emphasizing that AI functions as scalable pattern recognition rather than true clinical intelligence. In retina care, AI models are trained on large image datasets to detect referable DR, identify diabetic macular edema, and support image triage and workflow prioritization—providing probability outputs, not treatment decisions. Devisetty underscored that AI augments, but does not replace, retinal expertise and clinical judgment. She also highlighted emerging point-of-care wellness tools, including BioAge, CLAiR, and MyKidneyAI, which aim to expand AI-driven screening into broader cardiometabolic and kidney risk assessment for patients with diabetes.²

Cornea and refractive surgery: AI as clinical decision support

Lorenzo Cervantes, MD, of Connecticut Eye Specialists reviewed case-based examples illustrating AI in refractive screening and irregular cornea management. He framed AI as a spectrum—from rule-based algorithms to deep learning models—used to support risk stratification and treatment planning. In a refractive screening case, multimodal AI combined tomography and biomechanical data to assess subclinical ectasia risk. “AI converts complex data into objective risk scores for screening,” he noted

Looking ahead, Cervantes highlighted applications in keratoconus progression, infectious keratitis triage, and automated endothelial cell density analysis in eye banking. A deep learning system trained on more than 15,000 donor corneas can calculate endothelial cell density in under two seconds, reclaiming significant technician time. He cautioned, however, that many corneal AI models remain limited by homogeneous training datasets and lack of prospective validation.²

Retina research: AI as a catalyst for clinical trials

Hasenin Al-khersan, MD, of RCTX – Houston explored why AI has struggled to transform routine retina practice yet shows promise in research. He identified barriers including data silos, regulatory constraints, and workflow integration. “AI in research is a natural fit,” he explained, citing fewer regulatory hurdles and better economic alignment.

Applications include AI-enhanced prescreening to reduce high clinical trial screen-failure rates, enrichment of patient populations using predictive biomarkers, and synthetic control arms to decrease required enrollment. He noted that AI can detect subtle patterns invisible to clinicians, including predicting subject sex from fundus images with AUC 0.97.²

Cataract surgery & IOL power: Engineering precision with machine learning

Nambi Nallasamy, MD, of the University of Michigan Kellogg Eye Center presented a “whirlwind tour” of AI applications in ophthalmology, focusing on IOL power prediction and surgical analytics. His team demonstrated that machine learning–predicted postoperative anterior chamber depth (ACD) improves both traditional vergence formulas and ray-tracing methods for IOL calculations. “ML-powered postoperative ACD prediction improves both traditional [vergence] methods and newer [ray tracing, ML] methods,” Nallasamy reported.

The Nallasamy formula, trained on thousands of eyes, showed strong performance even when externally validated, underscoring the balance between generalizability and customization. He also introduced new evaluation metrics—MAEPI and Correct IOL Rate (CIR)—to address imbalanced real-world datasets. Beyond calculations, Nallasamy showcased AI-powered surgical video analysis, including instrument tracking (F1-score 0.953) and surgical phase segmentation (AUC 0.95), enabling objective assessment of surgical performance.²

Augmentation, not replacement

Across specialties, presenters converged on a central message: AI enhances pattern recognition, scalability, and efficiency, but physician oversight remains essential. From corneal ectasia screening and IOL prediction to retinal trial design and DR screening, Envision Summit 2026 made clear that AI’s greatest impact will depend not only on algorithmic accuracy, but on thoughtful integration, equity, and clinician trust.

References:

1. Chana R, Sikder S. Ophthalmic AI and Innovation. Presented at: Envision Summit 2026; February 12-16; Rio Grande, Puerto Rico.
2. Al-khersan H, Cervantes L, Devisetty L, Nallasamy N. Incorporating AI Into Your Clinical Practice. Presented at: Envision Summit 2026; February 12-16; Rio Grande, Puerto Rico.