Ultra-widefield Imaging Detection Rate in Identifying Peripheral Retinal Tears in Single versus Montage of Peripheral Sweeps

Abstract

Purpose: To compare the detection rate of orthogonal, directed peripheral steering, and auto-montaged images with ultra-widefield (UWF) imaging and the factors influencing the ability to identify retinal breaks. Design: Retrospective cohort study Methods: 376 treatment-naïve eyes (349 patients) that underwent laser retinopexy for retinal breaks between 2015-2021. Pre-treatment UWF orthogonal, peripheral steering, and auto-montage were cross-referenced to scleral-depressed examination to determine whether images successfully visualized all retinal breaks. Total retinal area visualized was divided by its optic disc area (pixels) to calculate relative retinal area (RRA). Potential associations were assessed by linear regression analysis. Results: 162 eyes (154 patients) met inclusion criteria. Orthogonal, peripheral steering, and auto-montage images showed detection rates of 47.5%, 90.7%, and 80.0% respectively. RRA increased from orthogonal versus montage by 34.7% 26.5% (mean SD), which increased detection rate by 90.8% (p=0.006). In linear probability models, vertical meridian tears decreased probability of identification in orthogonal, peripheral steering, and auto-montage by -26.6%, -86.2% and -68.7% respectively (p-values<0.001), and horizontal meridian tears increased the probability by 62.2%, 92.9%, and 85.5% (p-values<0.001). Tears posterior to the equator in orthogonal images increased the probability (91.4%,p<0.001). Artifacts such as lids/lashes, reflection, and faceguard decreased the probability in directed peripheral steering by -28.6%, -50.0%, and -66.7% respectively, (p=0.020, p=0.049 and p=0.016). Conclusions: Utilizing directed peripheral steering and auto-montage increases RRA and detection rate of identifying peripheral retinal breaks. Tears in horizontal meridians or posterior to the equator increase the probability of identification. Common UWF imaging artifacts can significantly limit the probability of identifying retinal tears.