Textural analysis in Meibomian gland image

Introduction: Meibomian gland is one of the anatomical eye structures to provide oily lipid layer to the anterior part of the eye. Textural analysis by using Mean Histogram Method was proposed as a method of estimating area of Meibomian Gland Loss (MGL). Objective: The main objective of this study was to objectively measure the meibomian gland image by using textural analysis. Methodology: This was a retrospective study consisted of 48 participants with readily available data of meibomian gland images captured by OCULUS Keratograph 5M. The data consisted of eye images from contact lens wearer (n=22) and non-contact lens wearer (n=26). The region of interests (ROIs) of total meibomian gland image were traced by one examiner by using ImageJ software. The other method adapted from previous project (Abdul Rahman N. F., 2015) named as Percentage Method. The method needed tracing to be done twice; total area of Meibomian gland and total area of MGL. The percent of MGL was calculated. Results: The repeatability of intra-grader was assessed by using intraclass correlation coefficients (ICC) that showed very high correlation based on Mean Histogram; right upper (ICC = 0.984, p < 0.001) and right lower eyelid (ICC = 0.983, p < 0.001). Spearman’s rho was used to test the correlation between Percentage Method and Mean Histogram Method. On right upper eyelid Spearman’s rho showed significant value (rho = -0.311, p = 0.031). Right lower eyelid showed very poor correlation (rho = -0.083, p = 0.573). The comparison of MGL between contact lens wearer (CW) and non-contact lens wearer (NCW) by using textural analysis method showed there was no statistical difference (P = 0.214 on upper eyelid, p = 0.634 on lower eyelid). Conclusion: Textural analysis based on Mean Histogram Method was shown repeatable. However, the measurement was not sensitive enough to discriminate between CW and NCW group