A Methodology for Glaucoma Disease Detection Using Deep Learning Techniques

Abstract

The advancement of computer technology and the needs of image processing is spreading in a wide range of applications. There are many techniques in image processing and one of the major techniques is Image classification. In the literature, we have reviewed many methods of machine learning used on Glaucoma pictures by different researchers. There are different machine learning algorithms include C4.5, the Naive Bayes Classifier, and Random Forest. These algorithms of machine learning cannot more reliably diagnose glaucoma disease. We have developed an architecture focused on the methodology of Deep Learning (DL), which is a Convolution Neural Network (CNN) for the classification of Glaucoma diseases. We used two different deep learning neural networks such as the Inception-V3 and the Vgg-16 Model for Glaucoma classification and identification purposes. We have obtained 508 Glaucoma images belonging to 25 groups from the Joint Shantou International Eye Center (JSIEC), Shantou City, Guangdong Province, China, Joint Shantou Foreign Eye Centre. Since uploading the images, we’ve increased the provided data set and rendered the 1563 training and testing data collection pictures. The downloaded data set is not labelled, so we wanted a named picture data set for our research in deep learning. But we have labelled both photos with the class name of the disease after the augmentation. We’ve also used two deep neural network models Inception-V3 and Vgg-16, which are supervised learning methods for classification arrangements. Such structures require operating processes that need to learn to use previous knowledge, make judgments about it, and fix it if any errors arise. We have used the Dropout: 0.5, Library: cv2, NumPy, Enjoinment API: Keras, TensorFlow, Loss Function: Cross-Entropy, Learning Rate: Adam, Fully Connected: SoftMax Activation Function with 2 Layer, Average Pooling: 4 Layer, Convolution: Tanh Activation Function with 2 Layer. Taking into consideration the success findings collected, it is shown that the pre-trained Inception-V3 model has the best classification accuracy with 90.01 percentage than Vgg-16 model which has an accuracy of 83.46 percent respectively.