Breast Cancer Detection using Histopathological Images
This work addresses early detection of breast cancer for pathologists and medical institutions, but it is incremental as it applies existing deep learning methods to a known medical imaging task.
The paper tackles breast cancer detection from histopathological images by training CNNs like VGG16 and ResNet on the BreakHis dataset to localize and classify cancerous regions, achieving results that emulate pathologists' actions for early diagnosis.
Cancer is one of the most common and fatal diseases in the world. Breast cancer affects one in every eight women and one in every eight hundred men. Hence, our prime target should be early detection of cancer because the early detection of cancer can be helpful to cure cancer effectively. Therefore, we propose a saliency detection system with the help of advanced deep learning techniques, such that the machine will be taught to emulate actions of pathologists for localization of diagnostically pertinent regions. We study identification of five diagnostic categories of breast cancer by training a CNN (VGG16, ResNet architecture). We have used BreakHis dataset to train our model. We focus on both detection and classification of cancerous regions in histopathology images. The diagnostically relevant regions are salient. The detection system will be available as an open source web application which can be used by pathologists and medical institutions.