An efficient hybrid methodology for an early detection of breast cancer in digital mammograms

Breast cancer continues to be a major health problem throughout the world impacting almost 2.1 million women each year. Delineation of breast cancer at an early stage can play a key role in the mitigation of the mortality rate in women. The presence of calcifications and suspicious mass regions (i.e. lesions) in digital mammograms are considered to be an early indicator of breast cancer. Hence, in this paper, the authors proposed an efficient hybrid methodology for the localization and identification of suspicious mass regions in digital mammograms. The proposed hybrid methodology is developed by integrating an efficient pixel-based low level pre-processing technique with a faster region-based convolution neural network (Faster R-CNN). These days, Faster R-CNN model is considered as a powerful object detection tool for medical image analysis. However, as a standalone tool, Faster R-CNN model offers several limitations with regards to breast cancer detection due to the mass regions being partially occluded by normal breast tissues, pectoral muscles, and noise that makes the mass detection a difficult and a challenging task. Therefore, to resolve the above issue, in this paper, an efficient mass detection methodology is proposed that involves the use of pixel based low-level preprocessing and Faster R-CNN approach. The performance of the proposed approach is evaluated in terms of various parameters such as sensitivity, accuracy, specificity, and area under the curve (AUC). The performance of the proposed model is also compared with other existing state of art algorithms such as Single Shot Detector (SSD), region-based fully convolutional network (R-FCN), and other deep learning based models. The proposed approach achieved the sensitivity, accuracy, specificity, and AUC score of 95.2%, 94.2%, 93.5%, and 0.983, respectively, which is quite satisfactory. © The Author(s), under exclusive licence to Springer-Verlag GmbH Germany, part of Springer Nature 2022.