ROPtool analysis of plus and pre-plus disease in narrow-field images: a multi-image quadrant-level approach

Quantitatively comparing weekly changes in retinal vascular characteristics of eyes eventually treated versus not treated for retinopathy of prematurity

“…This is consistent with a previous study that found no difference in the largest change per week in the following ROPtool indices between treatment versus nontreatment eyes: mean (ie, eye level) or maximum (ie, blood vessel level) TI, DI, or SAI. 14 Using quadrant-level methodology, 12 ROPtool could analyze (ie, trace at least 1 blood vessel for ≥1 disk diameter) 3-4 quadrants in 90% of our narrow-field retinal images (eSupplement 1). While an analysis of a subset of our data found that ROPtool could analyze 3-4 quadrants in 98% of the narrow-field retinal images, this subset only included one examination per infant (the latest imaging session prior to treatment for those eventually treated for ROP, and the imaging session with the most severe posterior pole disease that was closest to PMA of 36 for those not treated) with an average PMA of 35.6 weeks (range, 31.3-40.1; unpublished data) at the time those images were acquired.…”

“…While an analysis of a subset of our data found that ROPtool could analyze 3-4 quadrants in 98% of the narrow-field retinal images, this subset only included one examination per infant (the latest imaging session prior to treatment for those eventually treated for ROP, and the imaging session with the most severe posterior pole disease that was closest to PMA of 36 for those not treated) with an average PMA of 35.6 weeks (range, 31.3-40.1; unpublished data) at the time those images were acquired. 12 It was promising to see that ROPtool was able to analyze a high percentage of images when a wider range of PMAs were included (range, 30.9-51.4 weeks), which more realistically simulates the PMAs when ROP screening occurs.…”

“…ROPtool (v2.1.8) was used to trace images using the quadrant-level methodology. 12 Because ROPtool is a semiautomated computer program, user input was required to trace retinal images. Three investigators (two medical students [GJH, MCW] and one pediatric ophthalmology fellow [JCK]), all masked to clinical examination results, traced the images in a randomized order.…”

“…After it was determined which quadrants would be traced in which images, ROPtool was used to trace retinal vessels using the previously described quadrant-level methodology, 12 aiming to trace up to 2 major retinal vessels per quadrant. A "traceable vessel" was defined as one that ROPtool could trace for a length of ≥1 optic disk diameter.…”

“…8,10,11 More recently, a new technique combining quadrants from multiple narrow-field images of the same retina (ie, quadrant-level methodology) improved ROPtool's ability to trace vessels and showed high accuracy in identifying plus or pre-plus disease among images of varying quality. 12 While most studies have evaluated ROPtool's ability to identify the presence of plus or pre-plus disease at one point in time, 8,12 two studies have used ROPtool to analyze vascular changes over time. 13,14 One study found that as plus disease developed, changes in ROPtool tortuosity measures were sometimes very large, whereas changes in dilation measures were more subtle.…”