An ensemble-based approach for breast cancer identification using mammography

Abstract

Breast cancer is among the most common cancers in women worldwide; timely detection is vitally important for improving chances of survival. The present study examines an innovative machine learning technique for the diagnosis of breast cancer using the breast cancer Wisconsin (diagnostic) dataset from Kaggle. The dataset includes 569 instances, and each instance has 30 attributes derived from digitized fine needle aspiration (FNA) images of masses found in the breast. We will present an ensemble deep learning (DL) model fusing a convolutional neural network (CNN) and LRAlexNet architectures to increase the accuracy and robustness of this type of cancer diagnosis. CNN models are well-known for their power to capture spatial hierarchies in image data, and LRAlexNet is a specialized deep CNN that excels at image classification due to its depth and parameter optimization. In this work, we use the ability to extract features of CNNs along with the superior classification performance of LRAlexNet to distinguish between benign and malignant cancers. The model will be trained and validated on the curated breast imaging subset of the digital database for screening mammography (CBIS-DDSM) dataset, and performance will be evaluated using sensitivity, accuracy, specificity, and the area under the curve (AUC) for the receiver operating characteristic. The results show that the ensemble CNN-LRAlexNet model achieved superior accuracy for breast cancer prediction when compared to traditional machine learning methods.