Diagnosing Cancer Cells Using Histopathological Images with Deep Learning

Abstract

Early diagnosis of breast cancer for women can increase the survival opportunity with correct treatment in the clinic. Breast cancer can be diagnosed by detecting the malignancy of the cells of breast tissue. The microscopic images of breast cancer cells can be collected which can be used for detecting the presence of malignant (cancer) cells. The diagnosis process is tedious and the result may vary between pathologists. So, the Computer-Assisted Diagnosis (CAD) system is significant for improving diagnostic accuracy. The machine learning methods require export professional knowledge and experience to extract suitable features from histopathological images. A proposed deep learning-based approach can be used to classify breast cancer histopathological images in an automated way which utilizes different CNN models such as simple Convolutional Neural Network (CNN), dilated CNN, channel-wise separable CNN to extract image features and also uses Support Vector Machine (SVM) classifier and Softmax classifier to classify the histopathology images. This approach identifies the type of tumor and predicts the type of cancer cell as either benign (non-cancer) or malignant (cancer). This approach also identifies the best convolutional models which consume less time for training and also provide good training accuracy.