Binary differential evolution with self learning and deep neural network for breast cancer classification

Abstract

AbstractEarly classification of breast cancer helps to treat the patient effectively and increases the survival rate. The existing methods involve applying the feature selection methods and deep learning methods to improve the performance of the breast cancer classification. In this research, the binary differential evolution with self learning (BDE‐SL) and deep neural network (DNN) method is proposed to improve the performance of the breast cancer classification. The BDE‐SL feature selection method involves selecting the relevant features based on the measure of probability difference for each feature and non‐dominated sorting. The DNN method has the advantage which effectively analysis the non‐linear relationship among the selected features and output. The BI‐RADS MRI breast cancer dataset was applied to test the performance of the proposed method. The adaptive histogram equalization and region growing applied in the input images to enhance the image. The dual‐tree complex wavelet transform, gray‐level co‐occurrence matrix, and local directional ternary pattern were the feature extraction method used for the classification. This result shows that BDE‐SL with the DNN method has an accuracy of 99.12% and the existing convolutional neural network has 98.33% accuracy.