From Algorithm to Care: AI-Assisted Glaucoma Screening in Real-World Settings

Glaucoma screening is increasingly being discussed in the context of artificial intelligence (AI), particularly as imaging technologies become more accessible outside specialist clinics.The question, however, is not simply whether AI can identify a suspicious optic nerve head, but whether it can contribute to a screening pathway that works in real-world clinical settings.Fundus image-based AI has shown promising performance and is now being explored in community and primary eye care.The clinical value of these advances, however, depends on how effectively AI-assisted screening integrates into routine care pathways.Media opacity and poor-quality images can limit assessment, indiscriminate screening can generate unnecessary referrals, and many patients who are identified as requiring further evaluation may not complete the referral pathway.In glaucoma, the value of AI-assisted screening depends on its ability to facilitate appropriate clinical assessment and subsequent care for those identified as being at risk.AI is therefore best considered one part of a screening strategy rather than a replacement for clinical assessment.Risk-based selection, early recognition of ungradable images, and appropriate management of visually significant cataracts or other media opacity may make screening more efficient.Portable or tablet-based visual field testing could add functional assessment at the primary-care level, although its role remains to be established.The next phase of AI-based glaucoma screening should therefore focus less on improving performance metrics in isolation and more on understanding where the technology fits within existing eyecare services.The ultimate measure of success is not the number of suspicious optic nerve heads an algorithm identifies, but whether people with previously undetected glaucoma reach timely diagnosis, treatment, and ongoing care.