Artificial intelligence in managing retinal disease—current concepts and relevant aspects for health care providers

main topic Wien Med Wochenschr (2025) 175:143–152 https://doi.org/10.1007/s10354-024-01069-1 Artificial intelligence in managing retinal disease—current concepts and relevant aspects for health care providers Sophie Riedl · Klaudia Birner · Ursula Schmidt-Erfurth Received: 6 July 2024 / Accepted: 18 December 2024 / Published online: 24 February 2025 © The Author(s) 2025 Summary Given how the diagnosis and management of many ocular and, most specifically, retinal diseases heavily rely on various imaging modalities, the introduction of artificial intelligence (AI) into this field has been a logical, inevitable, and successful development in recent decades. The field of retinal diseases has practically become a showcase for the use of AI in medicine. In this article, after providing a short overview of the most relevant retinal diseases and their socioeconomic impact, we highlight various aspects of how AI can be applied in research, diagnosis, and disease management and how this is expected to alter patient flows, affecting also health care professionals beyond ophthalmologists. Keywords Optical coherence tomography · Artificial intelligence · Retinal disease · Imaging Retinal diseases Socioeconomic impact Global trends reveal that the numbers of patients affected by vision impairment and blindness are growing and are estimated to increase during the next 30 years [1]. Retinal diseases are the third most common reason for vision loss, following refractive errors and cataract [1, 2]. Globally, age-related macular degeneration (AMD) and diabetic retinopathy (DR) are most prevalent among retinal diseases, with AMD being the most common cause of irreversible vision loss in industrialized countries [2–4]. S. Riedl · K. Birner · Prof. U. Schmidt-Erfurth, MD () Department of Ophthalmology and Optometry, Laboratory of Ophthalmic Image Analysis, Medical University of Vienna, Währinger Gürtel 18–20, 1090 Vienna, Austria K Novel treatment strategies and continued technological advances have revolutionized retinal care in recent decades. Diagnostically, first and foremost, optical coherence tomography (OCT), offering non-invasive, cross-sectional, three-dimensional display of retinal tissue, has changed the management of numerous retinal diseases and is the imaging modality most commonly applied to the retina [5–7]. Therapeutically, intravitreal injections with anti-vascular endothelial growth factor (anti-VEGF) have opened up a novel therapeutic universe for several relevant exudative retinal diseases, including AMD, DR, and retinal vein occlusion (RVO) since their regulatory approval in 2006 [8–10]. Despite these groundbreaking advances, the chronic progressive nature of these conditions, requiring regular monitoring visits with re-injections every 4, 8, or 12 weeks, in addition to demographic shifts of rising patient numbers in a growing elderly population, lead to a high treatment burden for patients and health care systems [3, 11]. Additionally, in a recent breakthrough, novel treatment based on complement inhibition was approved for non-exudative AMD in the United States, which affects approximately 85% of AMD patients and might increase the treatment burden even further, as no treatment was available for this patient cohort until 2023 [12–14]. Overview of the most prevalent retinal diseases Degenerative Age-related macular degeneration AMD is estimated to affect 288 million patients worldwide in 2040 [3]. It is a multifactorial disease with genetics, age, smoking, and systemic disease as contributing factors for development and progression [15]. The pathophysiology behind AMD is not fully understood. The deposition of extracellular mate- Artificial intelligence in managing retinal disease—current concepts and relevant aspects for health care. . . 143 main topic rial and metabolic and vascular alterations due to chronic oxidative stress are believed to play a role [15, 16]. Drusen are the hallmark sign of AMD and histologically represent extracellular lipoprotein deposits located below the retinal pigment epithelium (RPE) [17]. Early and intermediate AMD (iAMD) stages are subclassified based on drusen size and so-called pigmentary changes of the macula and might progress to advanced AMD stages [17]. Patients with early and iAMD are usually asymptomatic, whereas mild symptoms including impairment of twilight vision and metamorphopsia occur in some cases [17]. Late AMD is categorized into neovascular (“wet”) AMD (nAMD) and non-exudative AMD, termed geographic atrophy (GA). Onset of nAMD presents with metamorphopsia, acute or subacute blurry vision, and scotoma, while GA is characterized by chronically enlarging lesions of atrophy, which lead to irreversible loss of visual acuity once the foveal center point is affected. GA lesion growth and the onset of vision-threatening symptoms show high interpatient variability [12, 17]. Early classifications of AMD were based on color fundus photography (CFP), while more recent nomenclature focuses on OCT imaging, which offers far more detailed visualization of pathologic features in all stages of the disease [18, 19]. In OCT, drusen, which are pathognomonic for early and iAMD, are elevations of the RPE. In nAMD, pathologic macular neovascularization (MNV) causes accumulation of fluid in the subretinal space (SRF) and intraretinal space (IRF). The vascular membrane corresponding to the MNV can be displayed in non-invasive OCT angiography (OCT-A) [18]. The underlying chronic progressive course of AMD leads to thinning and atrophy of outer retinal layers, which in its late stage presents as GA [19]. This process can readily be imaged by fundus autofluorescence (FAF) imaging and, again in more detail, by OCT [20]. Currently, there is no available treatment for early and iAMD. Initially, drastic vision loss in nAMD can be prevented by initiation of treatment with anti-VEGF intravitreal injections [21]. However, the chronic course of nAMD requires regular OCT monitoring and monthly, bimonthly, or quarterly reinjection, and long-term functional outcomes may still be unsatisfactory due to chronic disease progression, often leading to atrophy and scarring despite ongoing anti-VEGF therapy [22]. For GA, two novel therapeutics have been approved in the US since 2023 and are administered as monthly or bimonthly injections with the aim of slowing disease progression [13, 14]. Vascular diseases Diabetic retinopathy The global population of people affected by diabetes mellitus (DM) is estimated to reach 700 million patients in 2045 [23]. In DM, microangiopathy of retinal capillaries leads to ischemia and compensatory pro- 144 duction of proangiogenic stimulators and inhibitors, including VEGF [24]. Beside metabolic glucose control and systemic co-morbidities, disease duration remains a major risk factor for the development of DR [25]. In type 1 DM, 80% of patients will suffer from DR after 15 years of disease duration [26], while 84% (...truncated)