Automatic Cataract Detection and Classification Using Deep Neural Network

The function of the eyes, one of the vital organs of the human body, can be disrupted due to diseases such as cataracts. Age-related cloudiness and the formation of a waterproof eye layer are signs of cataracts. Each year, more than 120,000 new cataract cases are reported. Although early detection of cataracts can prevent total blindness, the preprocessing and feature extraction techniques used in current detection approaches are still limited. Therefore, this research aims to create an automatic system for cataract detection in fundus images by utilizing the Deep Neural Network architectural model. Four hundred ten fundus photos were used to train the model, divided into 329 for training and 81 for validation. The method is simple and consists of four main processes: data acquisition, pre-processing, model training using a neural network with a convolutional neural network, and testing of the cataract prediction model. Model performance is measured using a confusion matrix to compare expected and objective labels. Zero means no cancer, and one means cancer. From the results of the tests carried out, the highest accuracy of $\mathbf{9 9 \%}$ was achieved in the model testing results.