Purchase this article with an account.
Jalaiah Varikooty, Bruno Lay, Nancy Keir, Helene Burdin, Lyndon Jones, Trefford Simpson, Jessie Lemp; The relationship between clinical grading and objective image analysis of Lid Wiper Epitheliopathy. Invest. Ophthalmol. Vis. Sci. 2013;54(15):5460.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
To examine the relationship between clinically obtained grades of lid wiper epitheliopathy (LWE) and objective, automated image-processing derived metrics in a sample comprising subjects with a range of LWE.
In 22 subjects, the upper lid margin was stained with lissamine green (LG) dye using an optimized procedure to detect LWE. To reduce blur and improve depth of field, three images focused on the central, temporal and nasal regions of each upper lid were captured at 12x magnification through a digital camera (Canon EOS 60D) attached to a slit-lamp with constant camera settings maintained during the study. Using custom-built software with different user capabilities, each set of 3 images was automatically registered and processed and metrics characterizing the stained area were computed. The LWE was also graded by an investigator using a 0-3 scale (Korb et al. Eye Contact Lens, 2005). Mean difference, correlation, generalized linear model (GLM), and ROC analyses were undertaken using Statistica and R.
Median right and left eye investigator grades were 2.0 and interquartile ranges were 0.5 and 1.0 respectively. There were no differences between eyes (Wilcoxon matched pairs p=0.30). A number of image processing metrics were significantly associated with LWE grade: these included area, convexity and staining region thickness (Spearman rho all at least 0.61, all p<0.05). The GLMs showed significant relationships between LWE grades and the same imaging metrics. When LWE grades were binarized into approximately “Moderate” and “Severe” grades, ROC separation of these diagnostic groups was most effective with area, convexity and region thickness (Area under curve=0.96, 0.93, 0.88 respectively).
We were able to objectively characterize a number of aspects of LWE using this custom software. The most useful metrics in the study were staining area, convexity and thickness. These were strongly associated with the clinical grading using a diverse set of statistical tools. In addition to being more repeatable, objective measurements of LWE offer substantial numerical benefits and seem promising to measure even small changes as compared to quantized subjective LWE grades. This study was sponsored by Alcon Laboratories Inc.
This PDF is available to Subscribers Only