Application of artificial intelligence in diagnostics and surgery of keratoconus: a systematic overview

Abstract

Introduction. Artificial Intelligence is new theoretical approaches, methods, technologies and applied systems for modeling and extending human intelligence. In ophthalmology, artificial intelligence is one of the tools that help improve the efficiency of the treatment process through more accurate diagnostics, search for new biomarkers of diseases, automation of decision-making processes and assistance in other aspects of the physician‘s daily activities. The purpose of this review is to describe the currently available developments for the diagnosis and surgery of keratoconus in the field of artificial intelligence. Material and methods. Databases that were used for literature search included: Google and Google Scholar, PubMed, Embase, MEDLINE and Web of Science. Results. As a result of a search across all selected databases, as well as a selection of relevant studies, 75 articles were analyzed. Most of the studies that were selected for full-text analysis were the development of diagnostic algorithms. The most common classical machine learning methods were support vector machines method and random forest method. The most commonly used type of neural network is the convolutional neural network. 4 studies out of 75 reported the creation of a graphical interface for using the developed algorithm in a clinical environment. Conclusion. The accuracy of the algorithms that were obtained in the analyzed researches was basically more than 90%. It indicates the ability of machine learning models to solve complex clinical problems. Key words: artificial intelligence, machine learning, keratoconus, diagnostics