Utilization of pre-trained models of CNN in mammograms processing for the diagnosis of breast cancer
Meftahi Zahira Hanane, Merati Mejdeded · 2022
In this work, Pre-trained models of CNN on ImageNet were used and improved to design a perfect system with a lower error rate. The suggested system is essentially composed of two modules: (i) CADe for anomaly detection, that classifies images into normal or abnormal cases. (ii) CADx for identification, that classifies abnormal (cancerous) cases as malignant or benign. The proposed system was evaluated on two augmented datasets taken respectively from Digital Database for Screening Mammography (DDSM) and Invasive Ductal Carcinoma (IDC) databases.