Breast Cancer Detection Using Convolutional Neural Networks Model

Breast cancer is a significant disease that threatens people's health nowadays. A standard way to detect breast cancer is using radiology images by skilled physicians. However, this method causes problems in locating the cancerous area and intensive work in diagnosing and detecting histopathology images due to technical problems. The research results in the latest years can be divided into two directions, methods relying on machine learning or methods relying on deep learning. Due to the high dependency on the labeled data, we conducted two series of experiments that represent two reforming methods, analyzing a data set of a large number of breast cancer images to deal with these disadvantages. For the first series of experiments, we mutate the size of the training set using different models and compare the performances of the three models. For the second part, we do data augmentation on models to compare them before and after data augmentation and observe whether it makes these they achieve or reach the effect of the qualified model. Moreover, as the result shows, both our methods can reduce the workload in diagnosing breast cancer while maintaining or even improving the test accuracy, which benefits the follow-up work and development of the breast cancer diagnosis field.