Akademik Veri Yönetim Sistemi
Cornea, cilt.45, sa.3, ss.361-373, 2026 (SCI-Expanded, Scopus)
PURPOSE: The objective of this study is to enhance the efficiency and expediency of the corneal recipient selection process from extensive recipient lists by developing an interactive program that simulates the authorized operator with an artificial intelligence-based algorithm. METHODS: Taking into account the characteristics of real patients, 1000 distinct artificial corneal recipient candidates, each with 21 features, were generated as the recipient data set, and 50 distinct artificial donors, each with 5 features, were generated as the donor data set. Operators ranked the 20 most suitable corneal recipient candidates for each donor. These ranking lists, with the respective data sets, were employed during the training and inference stages of the multilayer perceptron used in machine learning. At the end of the training process, the top 20 recipient candidates identified by the operators for each donor were then compared with the top 20 recipient candidates identified by the system. RESULTS: The initial 11 candidates selected by the operator are endorsed by the system with a probability of at least 78%, whereas the initial 13 candidates selected by the operator are endorsed by the system with a probability of at least 67%. On average, the system identifies 17 out of 20 candidates, which corresponds to approximately 85% of all possible candidates. CONCLUSIONS: This study has resulted in the development of an interactive program that simulates the operator. This adaptable system, which does not impose specific selection criteria, can be trained in various eye banks worldwide and has the potential for extensive use.