Computer-Aided Detection of Malignant Mass in Mammogram Using U-Net Architecture

Abstract

X-ray screening image of the breast is called a mammogram and is used to diagnose breast illness like cancer. In this pioneer inquiry on screening mammograms, the normal and abnormal mammogram images have been analyzed. And an automatic CAD model is designed to aid radiologists by reducing manual empower. Computer-Aided Detection (CAD) is considered as a promising framework to diagnosis malignant mass; however, modern-day algorithms are hindered through excessive false predictions like calcium decomposition. Through this work, we implied a topical system for malignant mass recognition. As a substitute for locating candidate mass points and characterizing them, we discover anomaly at once at the entire mammogram slice by applying U-Net architecture. This is the deep learning approach where the discriminant features are identified without manual introspection. This prescribed model has experimented with the images acquired from the online Mammographic Image Analysis Society (MIAS) Digital Mammogram database and Texoma Medical Center (TMC) and it yields an 83% sensitivity rate for MIAS database and an 87% sensitive rate for the TMC database. The ultimate goal is to improve the prediction rate with discrimination between benign and malignant mass categories in well advance.