Precision in Breast Cancer Detection with Advanced Deep Learning for Lesion Localization and Classification

Abstract

Breast cancer is a global threat that mainly pertains to women. The need for detection at an early stage is crucial, but accurate detection of lesions due to breast cancer is still a huge challenge worldwide. This paper deals with the need for accurate detection and classification of lesions in mammogram images. A comprehensive study has been performed using the most advanced deep learning (DL) techniques. The study utilizes the DDSM dataset from the University of South Florida for mammographic image analysis. Among the models, the Faster R-CNN + ResNet-50 achieved 99.63% accuracy in lesion detection, while Inception V2 achieved 99.41% accuracy in tumor classification. These results highlight the effectiveness of the proposed approach in addressing challenges in breast cancer diagnosis.