Histopathological Image Analysis for Breast Cancer Detection Using Cubic SVM
Shiksha Singh, Rajesh Kumar
Abstract
Taking into consideration of world cancer report given by the World Health Organization (WHO) among women, breast cancer is the disease with the highest mortality rate worldwide. In Cancer disease, total of 25.2% of patients falls under the category of breast cancer. One of the main reasons behind the failure of saving cancer patients is due to latedetection and lacks of objective diagnosis in the type and level of cancer. Early detection of cancer has been improved due to evolutions in expert system and machine learning techniques with higher detection competence, for Computer-aided diagnosis. In this paper, histopathology-based feature has been taken into consideration for breast cancer detection and classification. The experimental analysis of the proposed approach has been done on publicly available dataset BreakHis. For experimental purpose we have tested K-Nearest neighbor (KNN), Random forest, and about six flavors of (SVM) Support Vector classification algorithms. The experimental result shows that proposed approach for detection and classification rate of breast cancer has been achieved maximum 92.3%accuracy with a cubic SVM classifier. The analysis of the results is verified with the help of classifier goodness parameters like accuracy, precision, recall, f-score, specificity, confusion matrix and ROC curve.