Computational Vascular Development Model Explaining Incidence of Notch in Retinopathy of Prematurity With...

Investigate retinopathy of prematurity (ROP) notch physiology, using ultra-widefield optical coherence tomography (UWF-OCT), and a computational model.

From June 2023 to October 2024, infants screened for ROP at the Oregon Health and Science University neonatal intensive care unit were imaged with an investigational 800 kilohertz (kHz), 140-degree, swept-source UWF-OCT. Eyes were included upon developing at least stage 1 ROP in zone I or II, with adequate image quality. Minimum retinal arc-length (min-RAL) to the vascular border were calculated. Then, a computational model simulated a modified-diffusion limited aggregation (DLA) system on a circular grid approximating the retina from ora to ora. Randomly released particles moved via Brownian motion until aggregating with the vascular network originating at the optic disc. An inferred central repulsive force operating by the inverse square law from the fovea was modeled. The simulation was terminated 50 times in zone I, and 50 in mid-zone II. A grader assessed both the simulations and the UWF-OCT images for a temporal notch.

Among 85 eyes (52 patients), notches appeared in 54 eyes (36 infants). Infants with a notch had lower gestational age (25.0 ± 2.0 vs. 27.4 ± 2.1 weeks), lower birthweight (533.6 ± 112 grams vs. 914 ± 359 grams), and shorter min-RAL (10.5 ± 2.6 vs. 12.9 ± 1.8 mm). The modified computational DLA model aligned predicting higher notch incidence in zone I (86% in simulation versus 83% in real eyes) than zone II (48% simulated versus 51% in real eyes).

This computational model introducing a foveal repulsive effect may explain varied retinal vasculogenesis patterns, like notch formation, and supports future predictive models.