Innovative utilization of ultra-wide-field fundus images and deep-learning algorithms for screening high-risk posterior polar cataract

Abstract

Purpose: Posterior capsular rupture (PCR) is a severe complication that occurs during cataract surgery. Patients with posterior polar cataract (PPC) are at a high risk of PCR. Mydriatic agents, despite notable side effects, enhance cataract posterior visibility during slit-lamp examination, aiding PPC identification. An ultra-wide-field (UWF) retinal imaging system, which does not need the aid of mydriatic agents, can visualize the projection (shadow) of a cataract onto the retina in fundus images. The relationship between PPC and the projected shadow remains unexplored. We hypothesized a cataract-shadow-projection theory and then validated it by developing a deep-learning algorithm which enables automatic and stable PPC screening using fundus images. Setting: Data were obtained from the Department of Ophthalmology at Far Eastern Memorial Hospital with permission from the hospital's Institutional Review Board. Design: Retrospective chart review data, including UWF fundus images. Methods: We developed a deep-learning algorithm to automatically detect PPC based on our cataract-shadow-projection theory. Retrospective data (n=546) with UWF fundus images were collected, and various model architectures and fields of view (FOVs) were tested for optimization. Results: The final model achieved 80% overall accuracy, with 88.2% sensitivity and 93.4% specificity in PPC screening on a clinical validation dataset (n=103). Conclusions: This study established a significant relationship between PPC and the projected shadow, enabling surgeons to identify potential PPC risks preoperatively and reduce the incidence of posterior capsular rupture during cataract surgery.