Breast Mass Tumor Classification from Mammograms Using Deep Learning

Breast cancer is widespread among females around the world. Detecting breast cancer in its earliest stages helps saving lives. Radiologists can decide if the depicted mammogram have cancer or not, but the miss rate is between 8% to 16%. Mammography is a popular screening for breast cancer masses. Because of mammography mostly manual nature, the tumors vary in form and boundary shape, also due to low signal-to-noise ratio, a notable number of breast masses are missed or misdiagnosed. In this thesis, we present two Convolutional Neural Network (CNN) models to classify pre-segmented mammogram mass tumor as benign or malignant. We test our approach on the Digital Database for Screening Mammography (DDSM) dataset. The classification accuracy reached 85.71%, with a recall rate of 87.3% and a precision rate of 85.7%.