Abstract
Purpose: :
Computer-assisted evaluation of disease severity in retinopathy of prematurity is an important diagnostic frontier, with implications for telemedicine as well as standardization of a subjective clinical evaluation. We have previously validated the ability of ROPtool to quantify separately tortuosity and dilation of posterior pole vessels. Here, our aim was to combine these parameters into a clinically meaningful measure and to assess the usefulness of this measure in evaluating posterior pole photographs.
Methods: :
"Tortuosity & Dilation Indices" (TI & DI) generated by ROPtool were compared to consensus grades of 2 authors (DKW & SFF) for 154 posterior pole vessels in 20 RetCam images. Using receiver operating characteristic (ROC) curves for both TI & DI, plus and pre-plus disease thresholds (or cut-off points) were chosen based on optimal sensitivity and specificity. All 80 quadrants were then scored based on the most dilated and the most tortuous major vessel in the quadrant, using a scale of 2 = plus, 1 = pre-plus, and 0 = neither plus nor pre-plus. Each quadrant’s "Plus Index" was calculated as the sum of the dilation and the tortuosity score, yielding a Plus Index of 0-4. These Plus Indices were compared to masked author consensus of plus disease, pre-plus disease, or neither.
Results: :
Compared with masked author consensus of pre-plus disease at the quadrant level, a Plus Index of 2 was 98.0% sensitive and 77.4% specific and a PI of 3 captured 91.8% of cases with no false positives (100% specificity). Compared with masked author consensus of plus disease, a PI of 4 had a sensitivity of 88.2% with 73.0% specificity, while a PI of 3 captured all cases (100% sensitive) with a 44% false positive rate.
Conclusions: :
The "Plus Index" represents a new, meaningful integration of ROPtool’s Tortuosity & Dilation Indices. It compares favorably to subjective ROP severity grades of experienced examiners in a pilot image set and should enhance ROPtool’s clinical functionality.
Keywords: retinopathy of prematurity • imaging/image analysis: clinical