An Ad Hoc Random Initialization Deep Neural Network Architecture for Discriminating Malignant Breast Cancer Lesions in Mammographic Images

<div>(a) Receiver operating characteristic (ROC) curves for a subsample of the architectures tested on the validation set (AUCs obtained on the validation set are shown in the legend). (b) ROC curve related to our best performing model (model 1: selected according to AUC on the validation set and model 2: selected according to <i>F</i>2 statistics on the validation set) when evaluated on the test set.</div>

Contrast Media & Molecular Imaging

fig5

Figure 5

Figure 5: An Ad Hoc Random Initialization Deep Neural Network Architecture for Discriminating Malignant Breast Cancer Lesions in Mammographic Images 