Osman Haji

Ophthalmology is entering a period of rapid and unprecedented transformation. Advances in artificial intelligence, high-resolution and home-based imaging, genomic medicine and service redesign are reshaping the clinical environment in ways that will fundamentally alter the role of the ophthalmologist. For today’s trainees (as well as tomorrow’s), understanding these shifts is an essential preparation for a career that will unfold during one of the most significant technological transitions the specialty has experienced.

Artificial intelligence (AI) represents perhaps the most visible of these developments. The pivotal prospective trial of IDx-DR demonstrated that a fully autonomous AI system could safely detect more-than-mild diabetic retinopathy in primary care with high sensitivity and specificity, marking the first FDA authorisation of an autonomous diagnostic system in any clinical specialty (1). Similar success has followed with systems such as EyeArt, which show robust performance in real-world diabetic retinopathy screening environments (2). Importantly, recent work has moved beyond diagnostic accuracy to examine the population-level effects of AI deployment. Studies of autonomous AI-enabled diabetic eye screening, for example, have shown significant increases in screening completion among groups historically underserved by traditional pathways, including children and young adults living with diabetes (3). These findings suggest that AI may increasingly function not merely as a diagnostic adjunct but as a key component capable of reshaping access to, and equity of care. Despite this progress, reviews of the literature remind us that real-world generalisability, dataset bias and the need for rigorous governance remain central challenges (4, 5, 6). The next decade is likely to witness not only the proliferation of AI tools within diabetic eye screening but also their integration into community triage, OCT interpretation and longitudinal risk prediction.

Alongside AI, the evolution of ophthalmic imaging is dramatically expanding the clinician’s diagnostic landscape. Deep learning has revolutionised the analysis of OCT, with contemporary systems capable of highly accurate classification of macular and optic nerve disease (7). Even more transformative has been the emergence of adaptive optics (AO) imaging, which allows near-cellular resolution of photoreceptors and other retinal structures in vivo (8). In inherited retinal diseases (IRDs), AO has refined genotype-phenotype correlations and enabled more sensitive detection of early degeneration, while multimodal imaging, which encompasses OCT, OCT angiography, widefield autofluorescence and AO, has become central to modern IRD phenotyping and trial eligibility (9). At the other end of the spectrum, interest in home-based imaging has grown rapidly. Prospective studies of the Notal Vision Home OCT system have demonstrated that patients with neovascular age-related macular degeneration (AMD) can successfully acquire daily retinal images at home, with scan quality sufficient for automated analysis and close agreement between AI and human graders (10). Such work suggests that chronic retinal diseases may eventually be monitored through a hybrid model of clinic-based examinations and remote structural imaging, reducing visit burden and enabling earlier detection of disease reactivation.

Perhaps the most exciting example of translational progress in ophthalmology has occurred in the field of genomics. The phase 3 trial of voretigene neparvovec for RPE65-mediated IRD demonstrated significant and clinically meaningful improvements in functional vision and light sensitivity compared with control subjects, establishing gene therapy as a legitimate therapeutic option (11). Subsequent studies, including long-term follow-up and data from Japanese cohorts, have confirmed sustained benefit with acceptable safety profiles (12). Meanwhile, the BRILLIANCE trial evaluating CRISPR-Cas9 editing for CEP290-associated Leber congenital amaurosis has provided early-phase evidence that in vivo gene editing can be performed safely in humans and can yield meaningful improvements in visual function in selected patients (13). As these therapies progress, ophthalmologists will increasingly need to understand variant interpretation, inheritance patterns, molecular mechanisms and the ethical considerations surrounding genomic testing. Precision medicine is also emerging in common diseases: polygenic risk scores for primary open-angle glaucoma have been shown to improve prediction of conversion from ocular hypertension to glaucoma and may in time inform screening strategies or personalised follow-up intervals (14).

Technological innovation is occurring in parallel with major structural changes in eye-care delivery. Pressures from ageing populations, chronic disease, and constrained clinical capacity have accelerated the adoption of virtual clinics, asynchronous review pathways and community diagnostic hubs. In glaucoma, virtual clinics, where technicians or optometrists collect standardised data for deferred consultant review, have been shown to be safe, effective and acceptable for patients with stable or early disease (15). Medical retina services have also adopted high-volume virtual models; one large diagnostic-hub-based clinic reported safe and efficient review of more than a thousand patients, with only a small fraction requiring urgent in-person assessment (16). There have also been evaluations of pop-up virtual clinics run by trained technicians in settings like shopping centres, which have demonstrated substantial reductions in waiting times while maintaining clinical safety, and are now being explored as scalable models for future NHS care (17). These pathways rely heavily on digital infrastructure, interprofessional collaboration and new models of clinical governance.

The configuration of the ophthalmic workforce is changing in tandem. Studies consistently highlight rising demand and significant workforce gaps, leading to widespread adoption of extended-role practitioners. The Ophthalmology Common Clinical Competency Framework has formalised advanced practice pathways for optometrists, nurses and orthoptists, enabling them to take on delegated clinical tasks (18). Intravitreal injection services provide a clear example of successful role expansion: nurse-delivered injection clinics across multiple centres have reported very low complication rates, including endophthalmitis rates comparable to medically delivered injections, along with high patient satisfaction (19, 20). These developments do not diminish the role of the ophthalmologist: rather, they shift its focus toward complex diagnosis, surgical and interventional expertise, and leadership of digitally augmented multidisciplinary systems.

For both current and aspiring trainees, this transitional moment carries significant implications. The ophthalmologists of the future will need to be adept at interpreting AI-generated outputs, understand the limitations of machine-learning models, participate in digital pathway design and navigate their ethical use. They will need fluency in multimodal and quantitative imaging, familiarity with home-monitoring technologies, and sufficient genomic literacy to counsel patients, liaise with genetic services and evaluate emerging therapies. Equally, they will be required to lead multidisciplinary teams, supervise extended-role practitioners and evaluate new models of care with a systems-thinking mindset. The traditional skills ophthalmologists need such as clinical examination, imaging interpretation and surgical skills remain foundational, but they will increasingly sit within a broader framework of digital, genomic and managerial fluency.

For junior ophthalmologists, preparing for this future involves cultivating expertise that aligns with these trajectories. Exposure to imaging-rich clinics, virtual glaucoma or retina pathways and IRD/genomics services will provide invaluable experience. Participation in research on AI, imaging biomarkers, genetic variants or service redesign offers opportunities to engage directly with the frontier of the specialty. Courses in basic data analysis, AI in healthcare, clinical genomics or leadership complement clinical training by equipping trainees with skills that are rarely taught but increasingly essential. Most importantly, trainees should focus not only on clinical competence but also insight, adaptability and the ability to contribute meaningfully to innovation and transformation.

Ophthalmology is entering an era defined by AI-supported diagnostics, ultrahigh-resolution and home-based imaging, gene and genome editing therapies and digitally reconfigured care pathways. The literature already documents how these technologies and models are influencing clinical outcomes today, and their integration into routine practice is likely to only accelerate. Trainees who embrace these developments, build the skills to interpret and deploy them safely, and cultivate leadership within multidisciplinary digital systems will not only remain relevant in a rapidly evolving landscape, but will influence the future of eye care itself.

References

1. Abràmoff MD, Lavin PT, Birch M, Shah N, Folk JC. Pivotal trial of an autonomous AI-based diagnostic system for detection of diabetic retinopathy in primary care offices. npj Digital Medicine [Internet]. 2018 Aug 28;1(1). Available from: https://www.nature.com/articles/s41746-018-0040-6
2. Vought R, Vought V, Shah M, Szirth B, Bhagat N. EyeArt artificial intelligence analysis of diabetic retinopathy in retinal screening events. International Ophthalmology [Internet]. 2023 Dec 1;43(12):4851–9. Available from: https://pubmed.ncbi.nlm.nih.gov/37847478/
3. Wolf RM, Channa R, Liu TYA, Zehra A, Bromberger L, Patel D, et al. Autonomous artificial intelligence increases screening and follow-up for diabetic retinopathy in youth: the ACCESS randomized control trial. Nature Communications [Internet]. 2024 Jan 11;15(1):421. Available from: https://www.nature.com/articles/s41467-023-44676-z
4. Abdi AS, Abdulazeez AM. A comprehensive review of deep learning in OCT image segmentation and classification. Medicine in Novel Technology and Devices. 2025 Dec;28:100396.
5. Muchuchuti S, Viriri S. Retinal Disease Detection Using Deep Learning Techniques: A Comprehensive Review. Journal of Imaging [Internet]. 2023 Apr 1;9(4):84. Available from: https://www.mdpi.com/2313-433X/9/4/84
6. Channa R, Wolf R, Abramoff MD. Autonomous Artificial Intelligence in Diabetic Retinopathy: From Algorithm to Clinical Application. Journal of Diabetes Science and Technology. 2020 Mar 4;193229682090990.
7. Lee CS, Baughman DM, Lee AY. Deep Learning Is Effective for Classifying Normal versus Age-Related Macular Degeneration OCT Images. Ophthalmology Retina. 2017 Jul;1(4):322–7.
8. Georgiou M, Kalitzeos A, Patterson EJ, Dubra A, Carroll J, Michaelides M. Adaptive optics imaging of inherited retinal diseases. British Journal of Ophthalmology. 2017 Nov 15;102(8):1028–35.
9. Daich Varela M, Esener B, Hashem SA, Cabral A, Georgiou M, Michaelides M. Structural evaluation in inherited retinal diseases. British Journal of Ophthalmology. 2021 May 12;105(12):1623–31.
10. Keenan TDL, Goldstein M, Goldenberg D, Zur D, Shulman S, Loewenstein A. Prospective, Longitudinal Pilot Study. Ophthalmology Science. 2021 Jun;1(2):100034.
11. Russell S, Bennett J, Wellman JA, Chung DC, Yu ZF, Tillman A, et al. Efficacy and safety of voretigene neparvovec (AAV2-hRPE65v2) in patients with RPE65-mediated inherited retinal dystrophy: a randomised, controlled, open-label, phase 3 trial. The Lancet [Internet]. 2017 Aug 26;390(10097):849–60. Available from: https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(17)31868-8/fulltext
12. Fujinami K, Akiyama K, Tsunoda K, Ito S, Seko N, Yamamoto S. Efficacy and Safety of Voretigene Neparvovec in RPE65-Retinopathy: Results of a Phase III Trial in Japan. Ophthalmology Science [Internet]. 2025 Jul;5(6):100876. Available from: https://pubmed.ncbi.nlm.nih.gov/40910102/
13. Pierce EA, Aleman TS, Jayasundera KT, Ashimatey BS, Kim K, Rashid A, et al. Gene Editing for CEP290-Associated Retinal Degeneration. The New England Journal of Medicine [Internet]. 2024 May 6;390(21). Available from: https://pubmed.ncbi.nlm.nih.gov/38709228/
14. Singh RK, Zhao Y, Elze T, Fingert J, Gordon M, Kass MA, et al. Polygenic Risk Score Improves Prediction of Primary Open Angle Glaucoma Onset in the Ocular Hypertension Treatment Study. medRxiv (Cold Spring Harbor Laboratory). 2023 Aug 16.
15. Mercer R, Alaghband P. The value of virtual glaucoma clinics: a review. Eye. 2024 Apr 8;38(10):1840–4.
16. Hanumunthadu D, Adan K, Tinkler K, Balaskas K, Hamilton R, Nicholson L. Outcomes following implementation of a high-volume medical retina virtual clinic utilising a diagnostic hub during COVID-19. Eye. 2021 Apr 6.
17. UCL. Technician-led eye clinics could lead to more timely NHS care [Internet]. UCL News. 2025 [cited 2025 Dec 10]. Available from: https://www.ucl.ac.uk/news/2025/jul/technician-led-eye-clinics-could-lead-more-timely-nhs-care
18. Greenwood V, Stanford P, Beddow C, Bowen M, Hingorani M. Changing practice for the non-medical ophthalmic hospital workforce in the UK—a snapshot survey. Eye. 2020 Sep 3.
19. DaCosta J, Hamilton R, Nago J, Mapani A, Kennedy E, Luckett T, et al. Implementation of a nurse-delivered intravitreal injection service. Eye [Internet]. 2014 Apr 4;28(6):734–40. Available from: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4058629/
20. Ahmed I, Maghsoudlou P, Hasan H, Abumattar A, Shah N. Safety and efficacy of nurse led intravitreal injection service with Precivia® injection assist device. European Journal of Ophthalmology. 2021 Nov 18;112067212110609.‌