IMAGE ANALYSIS BASED ON A NOVEL APPROACH OF ARTIFICIAL INTELLIGENCE FOR PREDICTION OF KERATOCONUS

Using a modeling system for early keratoconus diagnosis, one of the most challenging problems in ophthalmology, would provide a powerful tool for eye care professionals. We aimed to develop a novel model based on a chaotic system to extract the non-constant coefficients in an individual manner in every single case of keratoconus.

Bina Eye Hospital, Tehran, Iran.

Corneal properties were modeled using SIRIUS images. The designed model was implemented on two subgroups with 65 healthy cases and 32 keratoconus patients who participate in our study are between 23 to 74 years old (54 men and 43 women). To examine the accuracy of results, all cases are set into two subgroups (subgroup 1 and 2). The designed model was done in each subgroup and gathered data was analyzed.

The results of model analysis based on the chaotic system for divided subgroups in SIRIUS images reveal that significant sensitivity and specificity. Data results are 88% and 92% sensitivity for subgroup 1 and subgroup 2 respectively and specificity is 90% and 96% for subcategory 1 and 2 respectively (subgroup 1 p-value: 0.028 and subgroup 2 p-value:0.025). 

In this study, Keratoconus diagnosis with the help of a chaotic system with the characteristic of being exclusive to each patient is introduced by the (Iran)model. Promising results of accuracy and repeatability represent a novel approach in the field of keratoconus diagnosis.