Breast Cancer Detection and Localization Using a Novel Multimodal Approach

This article presents a novel multimodal Rényi cascade mask regional convolutional neural network (RCMRCNN) diagnostic system for breast cancer detection and localization. It consists of a reformed anisotropic diffusion high-boost filter (RADHBF) that is proposed to restore and enhance the multimodal database from the various multiplicative noises. Furthermore, multiscale feature learning is employed in the proposed method. It concatenates the single-scale features and provides accurate mapping of abnormalities. Apart from this, at the architectural level a cascade module of bounding box (BB) and a Rényi loss function have been proposed for the precise multi-instance localization of breast abnormalities. It is deployed on the NVIDIA Jetson Nano developer kit for the industrial applications. It exhibits the following values of accuracy, true positive rate (TPR), true negative rate (TNR), precision (PC), F-score (FS), balanced classification rate (BCR), Youden’s index (YI), Matthew’s correlation coefficient (MCC), Jaccard index (JC), and dice coefficient (DC): 98.50%, 99.00%, 98.02%, 0.98, 0.98, 0.98, 0.97, 0.97, 0.97, and 0.98, respectively.