Ensemble transfer learning networks for disease classification from retinal optical coherence tomography images

In this paper, we present a deep-learning-based approach for detection and classification of macular ailments from Optical Coherence Tomography (OCT) images of the retina. Retinal diseases, encompassing Choroidal Neovascularization (CNV), Diabetic Macular Edema (DME), and drusen, pose significant threats to vision health worldwide. Early and accurate detection of these diseases is crucial for timely intervention and treatment, which can significantly improve patient diagnosis outcomes. In this study, we propose a novel approach based on Ensemble Transfer Learning (ETL), employing two state-of-the-art deep learning models, DenseNet169 and InceptionV3, for the classification of retinal diseases. Transfer learning capitalizes on pre-trained models using extensive datasets to improve the efficacy such models in handling particular tasks. By fine-tuning these models on a curated dataset of retinal images, we achieve a remarkable 99.8% accuracy of disease classification. The ensemble of DenseNet169 and InceptionV3 further improves the resilience and universality of the classification system. The importance of this research lies in its potential to revolutionize early diagnosis and treatment of retinal diseases. By providing a rapid and accurate automated classification system, we empower healthcare practitioners to intervene at an early stage, potentially preventing irreversible vision loss. Additionally, this approach has the potential to alleviate the burden on healthcare systems by enabling efficient triaging of patients based on disease severity. The proposed Ensemble Learning (EL) model shows relatively good performance compared to other Convolutional Neural Network (CNN) models.