Plus disease: is it more than meets the ICROP? Insights about expert diagnosis from computer-based image analysis

Abstract

Purpose : Published definitions of “plus disease” in retinopathy of prematurity (ROP) reference arterial tortuosity and venous dilation within the posterior pole of a standard published photograph. One possible explanation for limited inter-expert reliability for plus disease diagnosis is that experts deviate from published written and pictoral definitions. The purpose of this study is to identify vascular features used by experts for diagnosis of plus disease through quantitative image analysis.

Methods : We developed a computer-based image analysis system (Imaging and Informatics in ROP, i-ROP), and trained the system to classify images compared to a reference standard diagnosis (RSD). The relationship of the performance of i-ROP as a function of the field of view (circular crops of 1-6 disc diameters [DD] radius) and vessel subtype (arteries only, veins only, or all vessels) was examined. The RSD was compared to the majority diagnosis of experts. A set of 77 digital fundus images was used to develop the i-ROP system. A subset of 73 images was independently classified by 11 ROP experts for validation. The primary outcome measure was the percentage accuracy of i-ROP system classification of plus disease with the RSD as a function of field-of-view and vessel type. Secondary outcome measures included the accuracy of the 11 experts compared to the RSD.

Results : Accuracy of plus disease diagnosis by the i-ROP system was highest (95%) when it incorporated vascular tortuosity from both arteries and veins, and with the widest field of view (6 disc diameter radius). This was comparable to the diagnostic accuracy of 11 expert clinicians (79–99%). Accuracy was <90% when using only arterial tortuosity (P=0.057), and <85% using a 2–3 disc diameter view similar to the standard published photograph (p= 0.004).

Conclusions : ROP experts appear to consider findings from beyond the 2-3 DD posterior retina when diagnosing plus disease, and consider tortuosity of both arteries and veins, in contrast to published definitions. It is feasible for a computer-based image analysis system to perform comparably to ROP experts, using manually segmented images.

This is an abstract that was submitted for the 2016 ARVO Annual Meeting, held in Seattle, Wash., May 1-5, 2016.