A Technical Review of Convolutional Neural Network-Based Mammographic Breast Cancer Diagnosis

Figure 6

The flow chart and an example of CNN performing as feature extractors. The flow chart highlights the information fusion which can be further divided into two approaches, feature fusion followed by a classifier or decision fusion of lesion malignancy predicted by using one or more classifiers. The example comes from [51] which develops a new CNN model and the model is pretrained on ImageNet. At last, the model fuses the prediction results from SVM classifiers which separately use 384 midlevel features and 2014 high-level features as its input.