Breast Cancer Classification Using Various CNN Models

Breast cancer is one of the leading causes of death among women globally. Early detection of tumors is possible through breast imaging, which is the first step in identifying abnormalities. Histopathology is a type of imaging techniques used to create a histopathological image of diseases in body tissues. This paper examines the use of Convolutional Neural Networks (CNNs) to classify benign and malignant tumors in the 400× and 100× BreaKHis datasets. We compare the results obtained using different types of convolutional neural networks models based on fastai to determine the highest accuracy level of classification among them, as well as the effect of image augmentation on the results. The first dataset showed that DenseNet 121 had the highest accuracy of 98%, followed by ResNet 34 (97.7%), and AlexNet 97%, respectively. The second dataset showed that ResNet 34 had the highest accuracy of 99.6%. It was also observed that enlarging the data before classifying had a negative impact on the results. Thus, the best solution for this kind of medical image classification is ResNet 34 without large magnification.