Algorithm of Caries Level Image Classification Using Multilayer Perceptron Based Texture Features

Abstract

A number of patients with untreated caries only seek treatment at late stages when serious complications might have already developed and can lead to significant acute and chronic conditions with high cost of treatment. The purpose of this research is to be able to find out the level of caries based on X ray images by using image processing and machine learning methods. The image processing algorithm namely Gray Level Co-occurrence Matrix (GLCM) has been used to extract texture features and Multilayer Perceptron (MLP) methods to classify the X ray caries images. Lavenberg Marquard and Backpropagation Bayesian Regularization are used in this study. The conclusion obtained in this study is that the algorithm of classification using Multilayer Perceptron (MLP) based texture features can classify dental caries images in four classes. The best performance result is achieved the training accuracy of 99.20% and the testing accuracy of 98.30% by using Lavenberg Marquardt (LM) model with hidden layer 10. In Backpropagation Bayesian Regularization (BR), the best results are found in hidden layer 10 as well (Training: 100%, Testing: 100%).