Ocular Disease Detection using Machine Learning

The prevention of visual impairment or blindness is largely dependent on early detection of ocular illnesses, which pose a serious worldwide health burden. Using a carefully selected dataset from Kaggle, this study investigates the use of machine learning techniques for the automated detection of ocular disorders. The dataset is made up of a wide range of retinal images that have been labeled with details on different eye disorders. The goal is to find a deep learning model that is reliable and precise enough to analyze retinal images and categorize them into several disease categories, such as glaucoma, macular degeneration, diabetic retinopathy, and normal conditions. A comparative analysis between different deep learning models like CNN, BCNN, VGG16 and VGG19 is performed in order to find the model that results in higher efficiency, that is, the model that best fits the dataset for multiocular disease detection. The idea of multi-disease detection was also proposed in the year 2021, which set a benchmark by gathering huge amount of relevant data for the study. It differentiated itself from the previous findings that where focused on predicting single ocular disease. Hence this research is a advancement in multi-disease detection where VGG16 proved to be the best fitting model for each label, with efficiency of $\mathbf{9 6 \%}$.