Convolutional Neural Networks for Breast Tumor Classification using Structured Features

Breast cancer is the second major reason for cancer death among women. Early diagnoses are incredibly crucial to increase the possibility of survival. Computer-aided detection has gained a lot of interest in the literature to help pathologists discriminate between malignant and benign tumors. Following remarkable successes on a wide range of diagnosis tasks, machine learning techniques attract substantial interest from medical researchers. In this paper, we propose a deep learning architecture to support breast tumors detection using structured features. First, we evaluated and compared the performance of multiple state-of-the-art machine learning approaches, including Support Vector Machine, Decision Tree, Logistic Regression, and Convolutional Neural Networks (CNN). Additionally, a combined ensemble model with three base models has been constructed to induce better generalization performance. We evaluated these approaches to automatically classify the tumors using a publicly available breast cancer dataset, the Wisconsin Carcinoma dataset (WBCD). Experimental results indicate that the highest classification accuracy (98%) belongs to the proposed CNN deep model classifier