Hybrid Integent Staging of Age-Related Macular Degeneration for Decision-Making on Patient Management Tactics

Treatment efficacy for age-related macular degeneration relies on early diagnosis and precise determination of the disease stage. This involves analyzing biomarkers in retinal images, which can be challenging when handling a large flow of patients and can compromise the quality of healthcare services. Clinical decision support systems offer a solution to this issue by employing intelligent algorithms to recognize biomarkers and specify the age-related macular degeneration stage through the analysis of retinal images. However, different stages of age-related macular degeneration may exhibit similar biomarkers, complicating the application of intelligent algorithms. This article presents a hybrid and hierarchical classification method for solving these problems. By leveraging the hybrid structure of the classifier, we can effectively manage issues commonly encountered with medical datasets, such as class imbalance and strong correlations between variables. The modifications to the intelligent algorithm proposed in this work for staging age-related macular degeneration resulted in an increase in average accuracy, sensitivity, and specificity of 20% compared to initial values. The Cohen’s Kappa coefficient, used for consistency estimation between the regression model and expert assessments of the intermediate class severity, was 0.708, indicating a high level of agreement.