Detection of breast cancer with ensemble learning using magnetic resonance imaging

Abstract

Despite notable progress in medicine along with technology, the deaths due to breast cancer are increasing steadily. This paper proposes a framework to aid the early detection of lesions in breast with magnetic resonance imaging (MRI). The work has been carried out using diffusion weighted imaging (DWI) and dynamic contrast enhanced-magnetic resonance imaging (DCE-MRI). Data augmentation has been incorporated to enlarge the data set collected from a reputed hospital. Deep learning has been implemented using the ensemble of convolutional neural network (CNN). Amongst the individual CNN models, the you only look once (YOLO) CNN yielded the highest performance with an accuracy of 93.4%, sensitivity of 93.44%, specificity of 93.33%, and F1-score of 93.44%. Using Hungarian optimization, appropriate selection of individual CNN architectures to form the ensemble of CNN was possible. The ensemble model enhanced performance with 95.87% accuracy, 95.08% sensitivity, 96.67% specificity, and F1-score of 95.87%.