Abstract
Purpose :
Retinopathy of Prematurity (ROP) is a leading cause of preventable blindness in children. ROPtool is a semi-automated computer program used to objectively measure retinal vascular characteristics in retinal images and has been validated to assess plus disease in RetCam and video indirect ophthalmoscopy images. Pictor is a hand-held, non-contact retinal camera. Our aim is to evaluate if ROPtool can accurately identify plus and pre-plus disease in a large set of Pictor images of varying qualities captured by non-physician health care workers (HCWs). While previous studies have used ROPtool to analyze single retinal images, we combined quadrant-level data from multiple Pictor images of a single retina to allow more major vessels to be analyzed.
Methods :
As part of an ongoing prospective study, HCWs obtained Pictor retinal images of infants at risk for ROP during routine ROP rounds. Imagers selected 1-3 images per eye, aiming to show vessels in all 4 posterior pole quadrants. A non-physician HCW analyzed these images with ROPtool. Six measures reflecting vessel tortuosity and dilation were obtained per quadrant: tortuosity index (TI), maximum tortuosity (MT), dilation index (DI), maximum dilation (MD), sum of adjusted indices (SAI), and tortuosity-weighted plus (TWP). The reference standard was the diagnosis at the time of the clinical exam. Receiver operating characteristic (ROC) curves were generated for the identification of plus and pre-plus disease using the second-largest quadrant value of each measure, because plus disease requires 2 quadrants to have sufficient vascular abnormality.
Results :
Of the 124 eyes imaged, 484 (98%) of 496 quadrants were analyzable by ROPtool, meaning they had at least one vessel traceable for at least one optic disc diameter in length. Overall, 92% of eyes had 4 analyzable quadrants, 6.5% had 3, and 1.6% had 2. For plus disease, area under the ROC curves (AUCs) were: TWP 0.97 = TI 0.97 = MT 0.97 > SAI 0.94 > DI 0.87 > MD 0.80. For pre-plus or plus disease, AUCs were: TWP 0.99 > MT 0.98 = TI 0.98 > SAI 0.97 > DI 0.87 > MD 0.80.
Conclusions :
Pictor retinal images of varying qualities of infants at risk for ROP can be analyzed with ROPtool with high accuracy for the identification of plus and pre-plus disease. It is feasible for non-ophthalmologist HCWs to capture retinal images and analyze them for the presence of plus disease, furthering the available workforce for ROP screening.
This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.