Classification of Eye Disorders Using Deep Learning and Machine Learning Models

Early diagnosis and screening of eye disorders are crucial for effective treatment. However, with the increasing prevalence of these conditions and a shortage of ophthalmic specialists, it is imperative to employ automated image evaluation methods to ensure consistent diagnoses and address this growing concern. Previous studies have focused on identifying eye conditions such as glaucoma, diabetic retinopathy, and cataracts using a single computer vision pipeline. In our research, we propose a hybrid approach that combines a classical neural network with four machine-learning models to classify these ophthalmic disorders. Our approach uses the pre-trained VGG16 model from the ImageNet dataset as a feature extractor. We then utilize the extracted features as input for the machine-learning models. To carry out this research, we harnessed a publicly available Kaggle database comprising images from various sources, including the Indian Diabetic Retinopathy Image Dataset, Oculur recognition, retinal datasets, and others. Finally, we conducted a comprehensive comparison of the results obtained with those from the top-rated deep learning models, to select the optimal model for diagnosing dry eye disease, which is our ultimate goal.