Ensemble Disease Learning Algorithm (EDL) for Retinal Diseases Detection

Abstract

Retinal diseases pose a significant threat to global eye health and require accurate and efficient diagnosis for effective treatment. Fundus images, obtained through non-invasive methods like fundus photography, provide valuable insights into the condition of the retina. In this study, fundus images are used to automatically detect and classify retinal illnesses using an Ensemble Disease Learning Algorithm (EDL). The EDL increases the robustness and accuracy of illness diagnosis by combining the predictive capability of several base classifiers. The diverse feature extraction techniques capture various aspects of retinal pathology, including optic disc morphology, blood vessel abnormalities, and lesion characteristics. These features are used to train a collection of base classifiers, such as Capsule Networks (CapsNets) for disease detection and disease classification using support vector machines (SVM). To create a robust ensemble approach, a weighted voting strategy that combines the decisions of individual base classifiers. The weights are determined through cross-validation and are adjusted to maximize overall accuracy. The proposed EDL algorithm enhances diagnostic accuracy and increases generalization by reducing the risk of overfitting. EDL on a comprehensive dataset comprising fundus images with various retinal diseases, including diabetic retinopathy (DR), age-related macular degeneration (ARMD), glaucoma, and typical cases. The experimental results demonstrate the superiority of the ensemble approach over individual classifiers, with an accuracy exceeding state-of-the-art methods. Finally, EDL presents a robust and accurate solution for the automated detection and classification of retinal diseases from fundus images. By harnessing the strengths of multiple base classifiers, we provide a reliable tool that can assist ophthalmologists in making well-informed decisions and thus contribute to early disease detection and timely intervention. This research signifies a substantial step towards enhancing the efficiency of retinal disease diagnosis and promoting better eye care on a global scale.