Dr. KRATI GUPTA

Dr. ROHIT SHETTY, Dr.Pooja Khamar, Dr.Gairik Kundu

Abstract

Aim: To study performance of AI based smartphone application in accurately identifying progressors of keratoconus (KC) & suggesting management.

Methods: 2500 scans were exported from Pentacam HR & classified into Stable & Progressing group. Keratometry parameters, KC indices & Zernike wavefront aberrations were given as features to AI. Machine learning was utilized to teach AI algorithm for management options like CXL, corneal transplant. Progression & management derived data was used to build smart-phone based application. Results: Random forest classifier-based AI model predicted disease progression with area under curve at 0.92, sensitivity & specificity at 0.8 & 0.87 respectively. Sensitivity & specificity of the KC app in successfully identifying progressors & suggesting a treatment was found to be 96.40%.

Conclusion: The KC app will help the general ophthalmologists in deciding the most suitable treatment option for their patient