An Expert System to Predict Eye Disorder Using Deep Convolutional Neural Network

Abstract

Glaucoma according to the W.H.O is one of the major causes of blindness worldwide. Due to its complexity and silent nature early detection of this disease makes it hard to detect. There have been several techniques over the years for classification which have shown significant improvement over the past decade or two. Some of the many classification models are SVM (support vector machine), KNN (K- Nearest Neighbors), Decision tree, Logistic Regression and ANN (Artificial Neural Network) back propagation. For this paper we would consider different procedure and method of early detection of the glaucoma disease using the MATLAB Deep Convolutional Neural Network (DCNN). The DCNN based expert system basically works like the human brain with input, neurons, hidden layers and output. For this project Fundus image of both healthy image and glaucoma image are collected with good lighting condition so that all hidden features can be identify. The Fundus image are then passed through different image processing method such as Grayscale, B&W, Complement, Robert, Resize and power Transform. The fundus is then passed through a texture feature extraction algorithm know as Deep Convolutional Neural Network (DCNN). The features gotten are Contrast, Correlation, energy, Homogeneity, Entropy, Mean, Standard deviation, Variance, skewness and Kurtosis. After the feature extraction the data are arrangement on a spreadsheet which serves as a means of record. Lastly, a deep convolutional neural network is written with one hidden layer, 16 input neuron and 2 output either healthy or not. The data are split into train and test dataset with 70% for training 15% validation and 15% for testing. Accuracy of detection was 92.78% with the execution time of 5.33s only depending on the number of iteration or epochs.