Clinical performance of an objective methodology to categorize tear film lipid layer patterns

Abstract

Purpose: To validate the performance of a new objective application designated iDEAS (Dry Eye Assessment System) to categorize different zones of lipid layer patterns (LLPs) in one image. Material and methods: Using the Tearscopeplus and a digital camera attached to a slit-lamp, 50 images were captured and analyzed by 4 experienced optometrists. In each image, the observers outlined tear film zones identified as a specific LLP. Further, the categorization by the 4 optometrists (observers 1, 2, 3 and 4) was compared with the automatic system included in iDEAS (5th observer). Results: In general, observer 3 was classified as worse than all observers (observers 1, 2, 4 and automatic application, Wilcoxon test, <0.05). The automatic system behaved similarly to the remaining three observers (observers 1, 2 and 4) showing differences only for Open Meshwork LLP when compared with observer 4 (Wilcoxon test, p=0.02). For the remaining two observers (observers 1 and 2) there were not found statistical differences (Wilcoxon test, >0.05). Furthermore, we obtained a set of photographs per LLP category for which all optometrists showed agreement by using the new tool. After examining them, we detected the more characteristic features of each LLP to enhance the description of the patterns implemented by Guillon. Conclusions: The automatic application in the iDEAS framework can provide zones similar to the annotations made by experienced optometrists. Thus, the manual process done by experts can be automated with the benefit of being unaffected by subjective factors.