Automatic Detection and Localization of Breast Cancer from Mammogram Imaging Modality using Modified Faster RCNN

The evolution of different imaging modalities leads to the study of the internal anatomy of the breast. It assists radiologist and physician in breast cancer detection. Breast cancer has claimed the lives of more women than any other type of cancer. With that in mind, this research proposes a strategy for detecting, classifying, and localizing breast cancer on mammogram type of imaging modality. Because Mammography of the breast muscles can detect breast cancer in its early stages and widely available. Therefore, this research work proposed a breast cancer detection framework based on Faster RCNN using mammogram images. It is help full to overcome the problem of miss classification and to save from unnecessary pain of biopsy. The proposed novel fully automatic CAD system can be used for helping the radiologist and doctor in detecting of the cancerous cells in the breast on mammography imaging modality. For being able to judge the proficiency of the suggested work. The CBIS-DDSM database for mammography is used. The result of our proposed method obtains accuracy of 94.12%, the true positive rate of 93.33%, and rate of true negatives is 94.74%, precision of 0.93, F-score of 0.93, BCR of 0.94 and Youden’s index of 0.88. The findings indicate that the proposed work provides a more effective technique for the earlier detection, classification, and localization of breast cancer. The result of qualitative and quantitative valuation show that proposed method is giving better results with respect to various performance evaluations metric such as accuracy, TPR, TNR, precision, F-score, BCR and Youden’s index.