Pick your player: AI or Human. Artificial Intelligence for Retinopathy of Prematurity Screening

Retinopathy of Prematurity (ROP) is a retinal disease affecting premature infants and remains one of the leading causes of childhood blindness world-wide. The early detection and treatment of severe disease is crucial to prevent vision loss, however screening is difficult due to lack of manpower and subjectivity of diagnosis. Our project aims to determine whether ROP.AI, an artificial intelligence algorithm developed to detect features of severe ROP, can perform as well as five expert ophthalmologists from across Australia. We found that ROP.AI was able to detect severe ROP with high sensitivity (84%) and negative predictive value (96%) compared to expert ophthalmologists. These findings demonstrate the vast potential ROP.AI holds to alleviate current limitations in ROP screening. Further algorithm training and prospective evaluation may see it be implemented as a decision support tool in the very near future.