EyeWorld Asia-Pacific March 2025 Issue

36 EyeWorld Asia-Pacific | March 2025 REFRACTIVE SURGERY ASIA-PACIFIC PERSPECTIVES Tanya TRINH, MD Cornea, Cataract and Refractive Surgeon, Mosman Eye Cente, 1A Effingham Street, Mosman, 2088 tanya.trinh@gmail.com AI In Refractive Surgery - What We Don’t Look For, We Miss Artificial intelligence (AI) continues to reshape refractive surgery by enhancing diagnostic precision, predicting patient outcomes, and tailoring treatment plans. While its transformative potential is undeniable, it is essential to critically examine the limitations and challenges associated with its adoption to ensure equitable and safe patient outcomes. Inherent Bias And Erroneous Assumptions By AI In Refractive Surgery AI systems are prone to inherent bias reflecting the limitations of their source data. There is significant growing concern about the paucity of appropriate data, for example in male versus female populations in almost every facet of medicine, leading to worse outcomes in female patients involving under or over-diagnosis or treatment. Ophthalmic population characteristics of the paediatric subset or the developing world are vulnerable populations where pre-existing source data may be constrained due to consent or accessibility issues. For example, effectively harnessing the power of AI to consider the risks of myopia progression among the paediatric population subset will requireAI to be trained on the appropriate age, ethnic and gender based data and not extrapolated from adults, due to the additional vulnerabilities of the developing visual system and the comparatively more vulnerable cornea of youth. Every effort must be made to train AI algorithms on diverse, representative datasets that accurately reflect the demographics and characteristics of the patient population being diagnosed or treated. And yet, these datasets may be challenging to obtain in resource challenged areas of the world, and have greater risk of potentially contributing to greater disparities in healthcare outcomes by inaccurate diagnoses and treatment recommendations. As refractive surgeons, we pride ourselves on practicing precision medicine; the optical challenges that we face demand it - and in turn, we must demand that our AI tools are based on the most accurate data possible and this means investing in obtaining the data as rapidly and as meticulously as possible. Without this, AI may inadvertently prioritize treatment strategies that are optimized for the majority population while overlooking the unique needs of marginalized groups. This can result in discrepancies in ectasia risk assessments or harmful postoperative visual outcomes for patients who fall outside the parameters of the majority dataset. Addressing this issue requires the

RkJQdWJsaXNoZXIy Njk2NTg0