Applications of Machine Learning in Glaucoma Diagnosis: A Systematic Review

Background and Aims: Early and accurate diagnosis of glaucoma is a major clinical challenge, as delays can lead to irreversible vision loss. Machine learning (ML) approaches, with their ability to simultaneously analyze complex data from multiple sources, hold significant potential to improve glaucoma diagnosis. The aim of this systematic review is to examine the applications of ML in glaucoma diagnosis and to evaluate their performance, limitations, and clinical applicability. Methods: This systematic review was conducted by searching five databases—PubMed, Scopus, Web of Science, ScienceDirect, and IEEE—from January ۱, ۲۰۱۰, to May ۳۰, ۲۰۲۳, using a structured search strategy. Following the PRISMA protocol for study selection, a total of ۲,۵۸۵ articles were identified. After applying inclusion and exclusion criteria, ۳۵ studies were selected for data extraction and qualitative assessment. Results: The included studies employed a wide range of ML techniques, including logistic regression, support vector machines (SVM), random forests, and deep learning algorithms. The most commonly used data sources were optical coherence tomography (OCT) measurements, followed by electronic health records and clinical features. ML models generally demonstrated high diagnostic performance, with area under the receiver operating characteristic curve (AUC) values often exceeding ۰.۹۰. However, direct comparison of results was challenging due to heterogeneity in datasets and study designs. Conclusion: ML-based clinical decision support systems show considerable promise in improving glaucoma diagnosis, achieving high diagnostic performance across various techniques and data sources. However, prospective validation across diverse populations is necessary to assess the real-world effectiveness and generalizability of ML in glaucoma detection

کلیدواژه ها:

Glaucoma ، CDSS ، Artificial Intelligence ، Machine Learning