Adopting Machine Learning to Automatically Identify a Suitable Surgery Type for Refractive Error Patients

Abstract

efractive error is a visual impairment that arises when the ocular anatomy hinders the proper focusing of light onto the retina, the light-sensitive tissue layer located at the posterior region of the eye. This condition poses difficulties in achieving clear vision. Refractive error stands as the prevailing kind of visual impairment. The objective of this study is to classify two surgical approaches utilized in the treatment of refractive defects. Two commonly performed refractive surgeries are Photo-Refractive Keratectomy (PRK) and Laser-Assisted In-Situ Keratomileusis (LASIK). Artificial Intelligence (AI) encompasses a specific branch known as Machine Learning (ML), which is the focal point of this investigation. ML is dedicated to the advancement and use of algorithms that possess the capacity to acquire knowledge from data and enhance their predictive capabilities without explicit programming. The present study employs sophisticated ML methods to classify different types of refractive defect surgeries using a dataset of 124 samples obtained from Al-Rabee Hospital in Iraq, specifically focusing on corneal topography data. Two ML approaches, namely K-Nearest Neighbor (KNN) and Artificial Neural Network (ANN), are employed to predict the kind of refractive defect surgery. The findings produced from the experiment demonstrated an accuracy rate of 90.32% for the KNN algorithm and a perfect accuracy rate of 100% for the ANN algorithm. Additionally, the KNN algorithm exhibited a sensitivity of 90% and a specificity of 90.54%. The study’s findings indicate that the ANN classifier outperforms the KNN classifier.