Lesion Detection and Classification for Breast Cancer Diagnosis Based on Deep CNNs from Digital Mammographic Data
Diksha Rajpal, Sumita Mishra, Anil Kumar · 2021
Breast cancer is the second most common cancer worldwide post lung cancer caused due to excessive growth of abnormal cells. The only way to restrict the effect of breast cancer is through diagnosis and screening of cancer in the early stages. The current diagnosis methods using deep learning has potential to improve breast cancer diagnosis in early stages. In this chapter, we use deep learning neural network for detection of breast cancer images as benign vs. cancer. Deep neural network is trained on public dataset MIAS with annotated images. The database consisting of only 322 images yields a low accuracy model; therefore, image augmentation is performed to obtain a larger training set which resulting in higher accuracy and better generalization capabilities. We then use unannotated labeled images collected from diagnostic center as test set to classify images as benign or cancer. This evaluation yields an accuracy of 92.74%, precision of 0.938, and recall of 0.89.