In recent years, significant progress has been made in the Meibography technique resulting from the use of advanced image analysis methods allowing a quantitative description of the Meibomian gland structures. Many objective measures of gland distortion were previously proposed allowing for user-independent classification of acquired gland images. However, due to the complicated nature of gland deformation, none of the single-valued parameters can fully describe the analyzed gland images. There is a need to increase the number of descriptive factors, selectively sensitive to different gland features. Here we show that global 2D Fourier transform analysis of infra-red gland images provides values of two new such parameters: mean gland frequency and anisotropy in gland periodicity. We show that their values correlate with gland dysfunction and can be used to automatically categorize the images into the three subjective classes (healthy, intermediate and unhealthy). We also demonstrated that classification performance can be improved by dimensionality reduction approach using principal component analysis.
Copyright © 2020. Published by Elsevier Inc.

Author