A Novel Deep Learning Based Approach for Breast Cancer Detection

This study examine and investigate possibility of employing machine vision techniques for Breast cancer identification. Initially, data preparation was performed for labelling followed by preprocessing for removing pectoral muscle and inherent noise. Digital Database for Screening Mammography (DDSM) an internationally available well received database, which contain 2620 cases of mammograms was incorporated for experimentation. VGG-16 was employed for feature extraction subsequently, SSD was incorporated for tumor detection. Once method was established results were compared with other published methods for validation. The results exhibit an accuracy of 96.2% on DDSM. Therefore, a method is proposed and developed and its effectiveness is demonstrated in term of breast cancer detection.