Perspectives on screening in retinopathy of prematurity: new algorithms and AI tools

Retinopathy of prematurity (ROP) is a vasoproliferative blinding disorder of the retina, unique to premature infants and a leading cause of preventable childhood blindness globally. Screening of premature babies aims to identify disease reaching a threshold for treatment needed in 6-10% of babies screened. Current screening criteria are based on birth weight (BW) and gestational age (GA) with an additional third criteria for babies deemed at risk by the neonatologist due to an unstable postnatal course or prolonged use of oxygen. These lack specificity and could potentially miss babies at risk. With increased survival rates of premature and extreme premature babies, ROP screening numbers are rising with associated clinical and economic burden. Several alternative algorithms have been proposed based on postnatal weight gain and in this review, the application and pitfalls of these will be discussed as well as the potential role of artificial intelligence in improving accuracy, efficiency, and equity in Australasia and low-middle-income countries.