Abstract
Vascular tortuosity (VT) is a critical biomarker of disease progression that informs the decision to treat ischemic retinal disorders, particularly retinopathy of prematurity (ROP). The murine oxygen-induced retinopathy (OIR) model is the most widely used model of ischemic retinopathy. Although VT has been described in OIR, its temporal dynamics have not been systematically defined. In this study, a semiautomated artificial intelligence–based (AI-based) pipeline was used to quantify VT throughout OIR. Retinal flat mounts from age-matched normoxic and OIR mice (P10–P56) underwent vessel segmentation using a generative adversarial network (GAN), and VT was quantified as a cumulative tortuosity index with the iROP-Assist algorithm. Concurrently, standard OIR outcomes of neovascularization (NV) and vaso-obliteration (VO) were quantified using the algorithm at http://oirseg.org/. NV peaked at P17 and resolved by P23, while VO regressed over a similar interval. VT peaked with NV at P17 but remained elevated through P56. These temporal changes mirror both the development of VT and its persistence after NV regression observed clinically in ROP. Collectively, these findings establish VT as a durable, quantifiable phenotype in OIR and expand the model’s utility beyond neovascular endpoints, providing a translational platform for investigating VT pathogenesis and evaluating the effects of therapeutic agents on VT.
Authors
Kyle V. Marra, Tomoya Murakami, Jimmy S. Chen, Edith Aguilar, Jacob Robinson, Maxwell Prenner, Richard Daneman, Martin Friedlander, Eric Nudleman
Figure 3
Quantification of neovascularization, vaso-obliteration, and vascular tortuosity in OIR and age-matched NOX mice.
