Estimation of Breast Cancer with a Combined Feature Selection Algorithm

Abstract

Image features are considered as a parametric factor that contains some of the specific information about the given image. In simple terms, a feature can be either a size or resolution or color information of an image. From the observed feature, a computer system can predict the nature of the image same as that of a human’s perception. In the beginning, the image processing algorithms utilized the features of the image only for the preprocessing and segmentation kinds of applications. An information regarding the noise ratio is considered for the preprocessing work to estimate the amount of smoothness needed to be given to the image. Similarly, the contrast difference or color difference features are widely employed by the segmentation algorithms. The proposed work aims to prove the efficacy of features on breast cancer image classification process using a multilayer perceptron algorithm. An experimental study is conducted on CBIS-DDSM dataset to estimate the importance of local and global features on breast cancer images.