Mammographic Mass Detection Based on Saliency with Deep Features
Yang Hu, Jie Li, Zhicheng Jiao · 2016
Breast cancer continues to be an attentional public health problem all over the world. Because of being indistinguishable from the surrounding normal tissues, automated mass detection is challenging in computer-aided detection (CAD) system. This paper presents a novel mass detection system for digital mammograms which integrates visual saliency model with deep learning techniques. After a visual saliency model applied to the pre-processed mammogram, a sliding window strategy is adopted to scan the breast area segmented from the whole mammogram. For each subimage produced by sliding window strategy, convolutional neural networks (CNN) features are extracted and propagated to a classifier to predict the class label. Large scale experiments on Digital Database for Screening Mammography (DDSM) demonstrate the effectiveness of the proposed mass detection method.