Abstract
Purpose :
Develop a new automated and objective methodology for assessing infrared meibography images to characterize meibomian glands structure and estimate meibomian gland drop out
Methods :
The Keratograph 5M (Oculus Optikgerate, Germany) was used to acquire infrared images of the tarsal conjunctiva. Ten subjects with different degree of meibomian gland dysfunction were measured. Bitmap files were exported and analyzed with a custom-built algorithm. First, the region of interest (i.e., tarsal conjunctiva) was selected using morphological image analysis and computing the local entropy of the image. In this region, a difference of Gaussians was performed to isolate the glands. After that, glands were labelled to separately assess their structure (length and width). Gland regularity is assessed by converting each gland from Cartesian to polar coordinates with respect to its center of mass and plotting the normalized values of angle vs radius. Finally, the area of meibomian glands drop out was defined as the percentage of the studied area not covered by glands.
Results :
According with the meiboscore proposed by Arita et al. (Ophtalmol., 2008) to estimate the percentage of meibomian glands drop out area, objective and automated grades (range from 0 to 3),were assigned to each image. According to this classification one subject had grade 0, five grade 1, three grade 2 and one grade 3. Table 1 shows the mean length and width of the glands and the percentage of drop out area for the different grades. Figure 1 shows an example of two eyelids one with regular glands (left) and one with tortuous glands (right). Red lines represent the limits were the glands are considered regular. The greater the values outside the red lines limits the more tortuous the glands are.
Conclusions :
The proposed method is able to grade the drop out area in an automated and objective way. In addition it provides information on glands structure (length, width and irregularity) which may be useful in meibomian glands dysfunction diagnostics and undertanding.
This is an abstract that was submitted for the 2018 ARVO Annual Meeting, held in Honolulu, Hawaii, April 29 - May 3, 2018.