Detection of Breast Cancer through Histopathological Images using Deep Learning

Abstract

One of the most prevalent occurrences of cancer amongst women is that of the breast cancer. The occurrence of this cancer in the female body leads to the impairment of various mental and physical related health issues. A large amount of complexities are however related to the presence of this disease which majorly includes the processing of images and its detection through image processing techniques. In addition to image processing techniques; various deep learning and transfer learning based techniques and algorithms can also be used in order to determine the same. For the purpose of implementation of the proposed research study; the author of the paper tends to utilize the Kaggle dataset which comprises of detailed information of breast cancer images which are obtained using the histopathological process. The dataset comprises of 800 cancer images and 250 healthy images which are undetermined with tumor. The entire process of defining the image takes place through the normalization technique using the concepts of histogram. Further, the author proposes to implement the working theory of CNN as a deep learning algorithm and Dense-Net-121 as transfer learning based algorithm. The transfer learning based algorithm however uses a pre-trained model to implement the same and further hyper-parameter tuning is also performed so that higher values of accuracy and precision can be obtained.