Classification of Caries X-Ray Images using Multilayer Perceptron Models Based Shape Features

Dental caries is one of the diseases that are often experienced by society, one way to detect it by taking pictures using Computed Radiography technology. The aim of this study was to develop a method of classifying dental caries imagery with the Hu Moment Invariant (HMI) and Multilayer Perceptron (MLP) feature extraction methods as an alternative to facilitate the detection of dental caries. The dental caries image used is a dental caries image for grade 1, class 2, class 3, class 4, with a total of 220 images of which 90% are as training data and 10% data testing. Hu Moment Invariant is used as a method of feature extraction and image classification using the Multilayer Perceptron (MLP) method. Classification is carried out with 2 classifier models namely Levenberg-Marquardt (LM), and Bayesian Regularization (BR) with a ratio of 3 types of Hidden Layer (HL) namely Hidden Layer 1, 5, and 10. The results of the analysis showed that the classification of dental caries imagery using HMI feature extraction and MLP classification will be obtained the best results when using the LM Hidden Layer 10 Model with the best training and testing accuracy results with a value of 96.1% and 98.3% and an average computing time between 1 to 14 seconds.