ABSTRACT
Breast cancer (BC) is the major cause of fatality among women in developing as well as under developing countries. Identifying the subtype of malignant (cancer) lesions can help to give proper treatment to cancer patients. The computer-aided diagnosis (CAD) is an automated and more accurate method for histopathological image classification. With the recent development in computer vision and deep learning, convolution neural networks (CNN) achieved enormous success in image classification and is widely used in medical image processing. To recognize the subtype of cancer automatically of the whole slide images (WSI), which is computationally impossible. The training on image patch from the images enables to design low-complexity CNN for feature extraction. The proposed Multi-Scale, Multi-Channel feature network (MuSCF-Net) using Resnet-Based Attention mechanism for breast histopathological image classification follows the knowledge sharing strategy by sharing learned features at each stream across the network and the attention mechanism. The proposed module achieved accuracy for different magnification factors (MF) (X40, X100, X200, and X400) but the superior accuracy, i.e., 97.57% for multi-class and 98.95% for binary at X200 MF. Performance measures like Accuracy, Recall, Precision, and Sensitivity, etc. parameters used for proposed module to classify histopathological breast images. We observed that MuSCF-Net model was better than existing models like VGG16, Xception, and ResNet152.
