Abstract
Purpose :
Lid wiper epitheliopathy (LWE) is a critical clinical marker to identify and track abnormal lid/cornea interaction. We proposed an automated grading algorithm using hue and saturation of LWE staining following manual selection of region of interest. Here we determined the repeatability of this semi-automated algorithm.
Methods :
31 images (15 upper and 16 lower eyelids) representative of LWE staining, observed after instillation of 20 microliters combination solution of 2% sodium fluorescein + 1% lissamine green, were analyzed. Images were processed using a custom MATLAB program that semi-automatically identifies the LWE area using hue and saturation values of the manually delineated lid region. LWE height (mm) was calculated as the average extent of staining perpendicular to the lid wiper curve minus the average Marx line height from images without LWE staining. Two independent examiners (E1 & E2) analyzed the images in a random order twice, with analysis sessions separated by more than a week. Bland-Altman analysis and Intraclass Correlation Coefficient (ICC) were performed to compare the results of height, width and area of the LWE staining.
Results :
Intra-examiner variability (E1) between analysis runs for height [-0.003 ± 0.011mm; 95% Limits of Agreement (LoA): -0.025 to 0.019], width [0.094 ± 0.478mm; 95% LoA: -0.843 to 1.031] and area [-0.003 ± 0.012mm2, 95% LoA: - 0.026 to 0.020] were calculated. Intra-examiner variability (E2) between analysis runs results for height [-0.001 ± 0.021mm; 95% LoA: -0.043 to 0.041], width [0.127 ± 0.670mm; 95% LoA: -1.187 to 1.440] and area [0.058 ± 0.291mm2, 95% LoA: -0.513 to 0.628] were calculated. The inter-examiner (E1 – E2) variability, using the average value of two analysis runs for each examiner, were calculated for height [0.001 ± 0.012mm; LoA: -0.022 to 0.023; ICC 0.999], width [-0.160 ± 0.395mm; LoA: -0.935 to 0.614; ICC 0.996] and area [0.024 ± 0.148mm2; LoA: -0.265 to 0.314; ICC 0.997].
Conclusions :
Inconsistency in subjective grading methods may prevent detection of changes in LWE staining between observers. This novel hue and saturation method results in highly repeatable measures of height, width and area of LWE staining after subjective delineation of the general lid region between examiners.
This abstract was presented at the 2022 ARVO Annual Meeting, held in Denver, CO, May 1-4, 2022, and virtually.