DETERMINATION OF POSITION OF INTRAOCULAR LENS IMPLANT BY DEEP LEARNING ALGORITHMS IN PATIENTS OF CATARACT SURGERY

Abstract

The current study was designed to predict the position of the intraocular lens in patients who have undergone cataract surgery using machine learning. A retrospective study was conducted in the Department of Ophthalmology at Lahore General Hospital from May 2022- May 2023. 150 patients undergoing cataract surgery by implanting intraocular lenses were selected for the study. The axial length, central corneal thickness, corneal curvature radius, and horizontal corneal diameter were noted during the surgery. The data set of 300 eyes was analyzed using the following techniques; regression model, regression trees, Support Vector Machine, and Gaussian Process Regression Models. The GPR technique with the exponential kernel gives perfect prediction compared to other algorithms. A mean root square of 0.185 mm was noted for postoperative AQD and 0.179 mm for the prediction of postoperative LEQ. The horizontal corneal diameter did not vary significantly in both models, and central corneal thickness was not associated with estimating postoperative AQD. Gaussian process regression is the best machine learning algorithm to predict the intraocular lens position in cataract surgery patients.