A quantitative method for assessing the quality of meibomian glands. (1st August 2016)

• Record Type: Journal Article
• Authors: Koprowski, Robert; Wilczyński, Sławomir; Olczyk, Paweł; Nowińska, Anna; Węglarz, Beata; Wylęgała, Edward
• Is Part Of: Computers in biology and medicine. Volume 75(2016)
• Page Start: 130
• Page End: 138
• Publication Date: 2016-08-01
• Subjects: Algorithm -- Biomedical -- Image processing -- Matlab -- Meibography; Medicine -- Data processing -- Periodicals; Biology -- Data processing -- Periodicals; 610.285
• DOI: 10.1016/j.compbiomed.2016.06.001
• Languages: English
• ISSNs: 0010-4825

Abstract

Abstract: Introduction: Meibomian gland dysfunction is a common cause of dry eye syndrome which can also lead to eyelid inflammation. Today, diagnostics of meibomian glands is not fully automatic yet and is based on a qualitative assessment made by an ophthalmologist. Therefore, this article proposes a new automatic analysis method which provides a quantitative assessment of meibomian gland dysfunction. Method: The new algorithm involves a sequence of operations: image acquisition (acquisition of data from OCULUS Keratograph® 5M); image pre-processing (image conversion to gray levels, median filtering, removal of uneven lighting, normalization); main image processing (binarization, morphological opening, labeling, Gaussian filtering, skeletonization, distance transform, watersheds). The algorithm was implemented in Matlab with Image Processing Toolbox (Matlab: Version 7.11.0.584, R2010b) on a PC running Windows 7 Professional, 64-bit with the Intel Core i7-4960X CPU @ 3.60 GHz. Results and conclusions: The algorithm described in this article has the following features: it is fully automatic, provides fully reproducible results – sensitivity of 99.3% and specificity of 97.5% in the diagnosis of meibomian glands, and is insensitive to parameter changes. The time of image analysis for a single subject does not exceed 0.5 s. Currently, the presented algorithm is tested in the Railway Hospital in Katowice, Poland. Highlights: The algorithm presented in this article is fully automatic – it does not require any operator intervention. The algorithm presented in this article provides fully reproducible results. The algorithm presented in this article provides sensitivity of 99.3% and specificity of 97.5%.