Artificial Intelligence May Contribute to Earlier Detection of AMD. How Does It Work in Practice?

AI in retinal examinations is not yet widely implemented in the Czech Republic. Devices using AI can be found in individual practices, and they are occasionally available in mobile clinics located in large cities during special events, such as World Sight Day. “I believe that given the potential savings in time and workload, this technology could become part of routine outpatient care in the near future,” says Dr. Martincová, expressing optimism about the future of AI-assisted examinations.

Accessibility and Speed as Key Advantages

Several devices worldwide operate with autonomous AI software; in the Czech Republic, the most commonly used system is Aireen. Its task is to detect the presence or absence of possible retinal pathology. It is certified at class IIb (EU-MDR) for diabetic retinopathy (2023) and age-related macular degeneration (2024).

“The examination takes place in a dark room where the patient places their head on the device support, and within seconds or at most a few minutes, the retina is photographed using a fundus camera,” explains Eliška Martincová. Pupil dilation is not necessary. The physician then reviews the captured images together with the AI output and determines the next steps. “If the finding is suspiciously positive, all such patients should be examined by an ophthalmologist,” she adds. The device thus triages patients into those with and without suspected pathology.

The aim of introducing AI-based technology is to relieve the burden on ophthalmology clinics and increase access to retinal examinations in primary care. This particularly concerns patients with diabetes, who may undergo screening at their primary care physician’s office without needing an additional specialist visit. This enables earlier detection in patients who might otherwise not reach an ophthalmologist in time. Compared with conventional procedures, the examination is also faster, allowing ophthalmologists to devote more time to complex cases.

Detection Even Before Clinical Manifestations

Support for early detection is particularly important in age-related macular degeneration (AMD), for which there is currently no nationwide screening program comparable to that for diabetic retinopathy. Implementing such screening could ease the workload of ophthalmologists and ensure earlier treatment. Patients often present at advanced stages associated with clinical symptoms –⁠ approximately one quarter of them only when both eyes are affected. In wet AMD, prognosis largely depends on how early therapy is initiated.

For home monitoring, the Amsler grid is commonly recommended, as it helps detect visual distortions and scotomas. However, AI can identify the disease even earlier. “The technology is capable of detecting early subtle retinal changes –⁠ drusen –⁠ which are asymptomatic and therefore not detectable using the Amsler grid,” explains the ophthalmologist.

According to general summaries, AI achieves 99.2% accuracy in distinguishing between eyes without pathology or with early signs of disease and those with intermediate or late AMD. In addition, it contributes to standardizing image evaluation (reducing interobserver variability). Some AI-based systems analyzing digital retinal images can even predict the risk of progression to late-stage disease within 1–2 years, for example the iPredict system. In the near future, AI may also assist in treatment decision-making, such as predicting treatment burden or selecting an optimal therapeutic regimen.

Editorial Team, Medscope.pro

Sources:
1. Crincoli E. et al. Artificial intelligence in age-related macular degeneration: state of the art and recent updates. BMC Ophthalmol 2024; 24 : 121, doi: 10.1186/s12886-024-03381-1.
2. Gao Y. et al. Recent advances in the application of artificial intelligence in age-related macular degeneration. BMJ Open Ophthalmol 2024; 9: e001903, doi: 10.1136/bmjophth-2024-001903.