Application of Faster-RCNN with Detectron2 for Effective Breast Tumor Detection in Mammography

Breast cancer is one of the leading causes of cancer-related deaths among women globally. Early and accurate detection of breast tumors through mammography plays a pivotal role in improving patient outcomes. Deep learning techniques, particularly CNNs, have demonstrated remarkable success in medical image analysis. In this study, we investigate the application of Faster R-CNN with detectron2, a state-of-the-art object detection framework, for the effective detection of breast tumors in mammographic images. The primary objective of our research is to develop an efficient breast tumor detection model that can assist radiologists in early diagnosis. In our study, we leveraged the power of transfer learning to adapt a pre-trained Faster R-CNN model to the specific task of breast tumor detection. Our experiments employ the INbreast dataset for training and evaluation, and we utilize the Detectron2 framework for model development. Fine-tuning the pre-trained Faster R-CNN model with the INbreast dataset allowed us to harness valuable features and knowledge from general object recognition tasks and adapt them to the intricacies of breast tumor detection. Our approach not only improves detection accuracy but also reduces the need for large-scale annotated medical image datasets, which can be challenging to obtain. Our results demonstrate the effectiveness of the transfer learning approach with Faster R-CNN in breast tumor detection. The model achieves high precision (AP50 = 0,93), making it a promising tool for assisting radiologists in their diagnostic tasks.