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

<div>Example images that are very difficult to classify: (a) image of a benign lesion that is falsely categorized as a malignant lesion (score 0.99992 from model 1 on a scale from 0 to 1); (b) image of a malignant lesion that is falsely categorized as a benign lesion (score .0133 from model 1 on a scale from 0 to 1).</div>

Contrast Media & Molecular Imaging

fig7

Figure 7

Figure 7: An Ad Hoc Random Initialization Deep Neural Network Architecture for Discriminating Malignant Breast Cancer Lesions in Mammographic Images 