Using convolutional neural networks and decision tree for Prediction of Breast Cancer

Abstract

Breast cancer is the leading cause of cancer-related mortality among women. Mammography surveillance plans have been proposed to decrease breast cancer mortality. However, the sensitivity of mammography is suboptimal, especially in women with “dense breasts.” In the analysis of breast images for cancer detection, lesion multiplicity and complexity may occur due to dense tissue interactions, making it challenging for radiologists to precisely detect and analyze masses. The present paper introduces a new computer-aided detection (CAD) system for breast cancer prediction. This system consists of three major steps: i) Segmentation strategies: The first strategy involves manually assigning the region of interest (ROI), while the second strategy uses a threshold- and region-based method, ii) Feature extraction: Features are extracted using a novel convolutional neural network (CNN), iii) Tumor classification: A decision tree is employed for classifying tumors. The performance of the proposed method is evaluated using the Digital Database for Screening Mammography (DDSM) dataset. The simulation results indicate that the proposed method outperforms existing methods, with an accuracy improvement of approximately 1.29%.