Author: Ivana Vrdoljak
First Examiner: Prof. Dr. Anna Nagl, Aalen University
Second Examiner: Damjan Žunić, M.Sc. Vision Science and Business (Optometry)
Delivery: October 31, 2025
Purpose of the Master Thesis:
This thesis explores the extent to which artificial intelligence (AI) can support, but not replace, human expertise in optometric care. It aims to evaluate the diagnostic capabilities and limitations of AI systems in detecting major ocular diseases—diabetic retinopathy, glaucoma, age-related macular degeneration, cataract, and keratoconus—and to examine ethical, regulatory, and professional implications of AI integration. The research further investigates how eye-care professionals perceive and use AI in clinical settings through a mixed-methods survey.
Methods:
A comprehensive literature review was conducted on peer-reviewed studies and meta-analyses concerning FDA- and CE-approved AI diagnostic systems. The analysis was supplemented by an original survey distributed to 81 optometrists and ophthalmologists worldwide. Quantitative data were analyzed in SPSS using descriptive statistics, Kruskal–Wallis tests, ordinal and multinomial logistic regressions, and correspondence analysis. Qualitative content analysis was performed on open-ended responses to identify common themes related to AI misdiagnosis, ethical concerns, and perceived professional impact.
Results:
AI systems such as IDx-DR, EyeArt, and Medios AI demonstrated high sensitivity (≥90%) and strong negative predictive value for detecting diabetic retinopathy, while AI in glaucoma, AMD, cataract, and keratoconus showed high potential but limited clinical generalizability. Among survey respondents, 45.7% reported occasional AI use, 11.1% regular use. 76.5% viewed AI as an assistive tool. Regular users rated AI’s diagnostic accuracy significantly higher than non-users (p = 0.010). Profession and years of experience influenced adoption of AI, with ophthalmologists and more experienced clinicians showing greater usage. Qualitative findings confirmed concerns about false positives, automation bias, and ethical accountability, while emphasizing that empathy, communication, and contextual interpretation remain uniquely human competencies.
Conclusions:
AI has become an indispensable adjunct in modern optometric diagnostics, capable of enhancing efficiency, consistency, and early disease detection. However, it cannot substitute for the holistic, ethical, and interpretative dimensions of human clinical judgment. The European AI Act and related medical-device regulations reinforce this principle through mandatory human oversight. The findings support a complementary model in which optometrists and ophthalmologists collaborate with AI to deliver safer, fairer, and more comprehensive eye care.
Keywords: Artificial intelligence, Optometry, Ophthalmology, Diagnostic accuracy, Diabetic retinopathy, Glaucoma, Ethical oversight, Human expertise, EU AI Act, Automation bias