A Hybrid Technique for Classifying Ocular Diseases Using Deep Learning and Machine Learning Algorithms

Ocular problems, including glaucoma and retinal detachment, are caused by different factors such as diabetes. They cause a range of severe symptoms of blindness around the world. Early detection is challenging due to the complexity of time demands and potential inaccuracies in manual diagnosis. This research proposes a hybrid deep learning approach that utilizes feature extraction using convolutional neural networks (CNN) for classification via traditional machine learning techniques. The Ocular Diseases Intelligent Recognition (ODIR), a dataset containing 10,000 ocular images from 5,000 persons, was utilized for performance evaluation. This research seeks to deliver precise diagnoses of ocular disorders using deep learning methodologies, thereby enhancing healthcare quality, especially ocular health. The findings show that the best accuracy (97.34%) for eight-class classification was attained using a proposed CNN for feature extraction and employing a multi-support vector machine (MSVM) as the classifier to score a high accuracy. The results indicate that the suggested model may enhance accuracy in diagnosing eye diseases while addressing challenges in multi-class classification.