Keratoconus is an eye disease of 'deformation of corneal curvature' caused due to 'non-inflammatory progressive thinning' resulting into loss of elasticity in cornea and protrudes a cone shape formation that ultimately reduces visual acuity. For many years, researchers have worked towards accurate detection of keratoconus (KCN) as it is essential checkup before any refractive surgery demanding quick as well as precise clinical diagnosis and treatments of keratoconus prior to LASIK. In our study, we have firstly derived two variants of the original corneal topographies namely 'images with edges' and 'images with edges-and-mask', as data sets. The deep neural network techniques such as Artificial Neural Network (ANN), Convolutional Neural Network (CNN) and pertained VGG16 model are applied on original 'corneal topographies' as well as on the two of its variants and the results obtained are presented. Keywords: Keratoconus Á Corneal topography Á ATLAS 9000 Á ANN Á CNN Á VGG16 Á Canny edge detection Á Edges with mask

Keratoconus is an irreversible and progressive deformation of cornea and hence it is critical to be detected in its early stages. Moreover, since, an eye with advance stage of Keratoconus shall not be operated for refractive eye surgery, further makes it important to diagnose Keratoconus in early stages itself for sake of effective treatment without subsequent complications. Hence the researchers have been toiling hard to be able to detect the disorder in its early stage, since decades, with the help of Artificial Intelligence, Neural Networks, Machine learning and Deep learning etc. These advance techniques are helping in analyzing and classifying Keratoconus eye’s data for predicting it in advance. Decades of quantitative data could be digitalized with the advancement of eye screening machines used for clinical examination of eye functionalities. Here, in this study, we have analyzed and tried to summarize the role of respective technologies in identification and classification of Keratoconus. Further the collection of eye data gathered from various topographers through ophthalmic screening machines is also presented here-in along with the timeline of use of neural network techniques for detecting the Keratoconus eyes. The objective of the study is to identify the appropriate neural network technique that could identify the disease in its early stage using the corneal topographical images as data.

No abstract