<![CDATA[This study aims to detect cataract disease in retinal images obtained from retinal scans using a Siamese Convolutional Neural Network (SCNN) model. The Dataset used consists of retinal images captured directly from patients using an ophthalmoscope and collected from various sources. The Dataset includes images showing cataract disease as well as images from normal eyes. These images are then converted to grayscale to facilitate feature extraction, enhanced using Histogram equalization, then paired and labeled for use in the SCNN model. The data is divided into training, validation, and test sets for model training and testing. Training is conducted using the Tensorflow Keras framework with the SCNN model. Testing is performed in three stages: first, system performance testing with the use of Histogram equalization; second, learning parameter variations; and third, testing the identification of the SCNN model to detect cataracts. The best model is obtained using 100 epochs, RMSProp optimizer, and Binary Crossentropy loss function, achieving a test accuracy of 91.25% with a prediction time of 0.1113 seconds. Thus, an SCNN model has been successfully developed and implemented to detect cataract disease in the eye. Keywords— Cataract, Histogram equalization, Siamese Convolutional Neural Network (SCNN)]]>
Copyrights © 2024
