Abstract
Purpose:
Appropriate screening and treatment can reduce the risk of vision loss from retinopathy of prematurity (ROP). To minimize subjectivity in screening for plus disease in ROP, computer programs such as ROPtool have been created to analyze images by quantifying vessel dilation and tortuosity. ROPtool has limited ability to analyze images of poor quality or contrast. Pictor is a portable, non-contact camera that can capture high-quality retinal images of infants at risk for ROP. Our aim was (1) to compare ROPtool’s ability to analyze retinal images acquired by Pictor versus video indirect ophthalmoscopy (VIO) and (2) to assess the traceability of Pictor images by ROPtool as an imager acquired experience using Pictor.
Methods:
We included all infants who had Pictor imaging obtained by a single imager and VIO images taken during routine ROP rounds from May 2013-Oct 2014. We enhanced our sample with images of pre-plus and plus disease by choosing images from the exam immediately preceding treatment, for those who were treated. Otherwise, images were selected first by most severe posterior pole disease, then by infant postmenstrual age closest to 36 weeks. We selected one VIO and one Pictor still image from one eye per infant from the same examination. ROPtool was used to analyze each image. An image was considered “traceable” if ROPtool could trace at least one vessel in each quadrant for a length of at least one optic disc diameter. We calculated traceability over the entire study period, and we also calculated the traceability of images for the first and second half of the study.
Results:
We analyzed 16 Pictor and 16 VIO images from 14 infants. Image traceability for the study period was 93% (13/14) for Pictor and 71% (10/14) for VIO images (p=0.1). Traceability of Pictor images was 89% (8/9) for the first half and 86% (6/7) for the second half of the study (p=0.5).
Conclusions:
Compared with retinal images acquired by VIO, there is a trend toward Pictor images being more traceable by ROPtool. Even images captured early on in an imager’s experience of using the Pictor were able to be traced by ROPtool. The quality of images captured with Pictor are well-suited for image analysis, bringing us closer to developing a portable and non-contact imaging system to complement automated plus disease diagnosis.