Breast Cancer Identification and Classification Using Contextual Deep Learning Technique in Histopathological Images
G N Keshava Murthy, Piyush Kumar Pareek, C P Nayana, I M Ramya, Keerthi Kumar M, Rashmi Priya · 2023
Breast aggressive ductal carcinoma (IDC) are both forms of ductal carcinoma, and the pathological distinction between the two is crucial for selecting the most effective course of treatment and predicting the most favourable clinical results. Since there is a finite number of pathologists who can use microscopes for conventional diagnosis, it is imperative that novel approaches be developed that can promptly and precisely identify huge statistics of histopathological specimens. It would be tremendously helpful for routine pathological diagnosis if computational pathology tools existed to aid pathologists in the detection and classification whole slide images (WSIs). In this study, we contemporary a deep and broad contextual convolutional neural network (CCNN) for histopathology image categorization, which outperforms state-of-the-art deep networks. By efficiently utilising local spatio-spectral correlations of adjacent separate vectors, the suggested CCNN can discover previously undiscovered local contextual interactions. One step in the proposed CNN pipeline is a filter bank, which allows for the simultaneous utilisation of data. Once the multi-scale filter bank's spatial and spectral feature maps have been created, they are integrated to create a joint spatio-spectral feature map. When a fully convolutional network is fed the joint feature map containing the breast cancer image's rich spectral and spatial features, it may make predictions about the labels that should be assigned to each individual pixel vector. The suggested method is assessed using dataset, and a comparison of its performance to that of state-of-the-art techniques is provided.