A New Dawn in Eye Analysis
Artificial intelligence is beginning to transform how we can bring together and interpret the wealth of information collected during an eye examination.
A major recent development in my practice has been the introduction of AI-assisted ophthalmic analysis, and this has now progressed further with the development of a dedicated Ophthalmic AI system, configured specifically for consultant-level eye care and with particular emphasis on Medical Retina.
Anonymised retinal photographs from the Topcon NW500, Zeiss Cirrus OCT scans, optic nerve analysis, visual fields, eye pressures, vision measurements and relevant clinical history can be assessed together rather than simply being considered as a series of separate tests.
The system is designed to act as an additional clinical second opinion. It analyses the available evidence independently, considers alternative diagnoses, looks for inconsistencies and can challenge an initial interpretation rather than simply agreeing with it.
One particularly useful development is automated drusen analysis in age-related macular degeneration (AMD). From suitable retinal photographs, AI can assist in assessing the approximate number, size, type and distribution of drusen, including whether they are predominantly small, medium or large, discrete or confluent, and whether there are associated pigmentary or retinal pigment epithelial changes.
This provides a more detailed assessment of the AMD phenotype than simply recording whether drusen are present. When previous photographs and OCT scans are available, changes in drusen, retinal structure and areas of atrophy can also be compared over time.
The same longitudinal approach can be applied to many other conditions. AI can help compare scans taken months or years apart, assess changes in retinal fluid or thickness, examine the progression of geographic atrophy, review the response to anti-VEGF treatment, analyse diabetic retinopathy and previous retinal laser treatment, and identify subtle changes that may otherwise be difficult to quantify.
The real advance is therefore not AI looking at a single photograph or OCT scan. It is its ability to bring together many different pieces of clinical information, analyse how they relate to one another and compare them over time.
Where appropriate, the system can also review current published research and treatment evidence, providing another layer of information when considering newer or evolving treatments.
Importantly, AI does not diagnose or treat patients independently and does not replace the ophthalmologist. All AI-generated observations are critically reviewed and interpreted by Mr Lee, alongside the original images, clinical examination, previous records and his own experience and judgement as a Medical Retina specialist.
AI measurements from photographs, including drusen counts and size estimates, should also be regarded as assisted estimates rather than laboratory-grade measurements, particularly where an image does not contain a calibrated scale.
Although assembling and reviewing this amount of information remains time-consuming, it is allowing a substantially deeper and more individualised assessment of complex eye conditions. The combination of high-quality ophthalmic imaging, long-term clinical records, specialist experience and carefully controlled AI analysis offers an exciting new approach to monitoring eye disease and supporting clinical decision-making.
This takes the consultation to a new level.