Dhoriva Urwatul Wutsqa, Retno Subekti, Rosita Kusumawati
In this paper, we develop a fuzzy neural network model for classifying cervical cancer. A fuzzy neural network is a specific feed forward neural network which processes fuzzy variables. The fuzzy neural network model gives benefits, since the fuzzy logic accommodates cognitive uncertainty, and neural network allows learning and generalization. The cervical cancer is classified using colposcopy images. We also propose image preprocessing of intensity adjustment and median filter to enhance the contrast and to shrink the noise of the image. The parameters extracted from the image by the grey level co-occurrence matrix method are designed as the original inputs of the fuzzy neural network model. These inputs are in crisp forms, so they require to be changed in fuzzy forms. In this study, we use trapezoidal fuzzy number and OR operation. The fuzzy neural network is implemented to the original colposcopy images, to the colposcopy images with the intensity adjustment, with the median filter, and with the combination of both. The results demonstrate that the fuzzy neural network models have the same performance to all types of images on training data and deliver the best performance on the images with the combination of intensity adjustment and median filter on testing data. © 2019 International Center for Scientific Research and Studies.
Department of Statistics, Faculty of Mathematics and Natural Sciences, Universitas Negeri Yogyakarta, Indonesia
