Abstract
Breast cancer is one of the deadliest diseases worldwide among women. Early diagnosis and proper treatment can save many lives. Breast image analysis is a popular method for detecting breast cancer. Computer-aided diagnosis of breast images helps radiologists do the task more efficiently and appropriately. Histopathological image analysis is an important diagnostic method for breast cancer, which is basically microscopic imaging of breast tissue. In this work, we developed a deep learning-based method to classify breast cancer using histopathological images. We propose a patch-classification model to classify the image patches, where we divide the images into patches and pre-process these patches with stain normalization, regularization, and augmentation methods. We use machine-learning-based classifiers and ensembling methods to classify the image patches into four categories: normal, benign, in situ, and invasive. Next, we use the patch information from this model to classify the images into two classes (cancerous and non-cancerous) and four other classes (normal, benign, in situ, and invasive). We introduce a model to utilize the 2-class classification probabilities and classify the images into a 4-class classification. The proposed method yields promising results and achieves a classification accuracy of 97.50% for 4-class image classification and 98.6% for 2-class image classification on the ICIAR BACH dataset.
Reference46 articles.
1. Epidemiology of breast cancer in Indian women;Malvia;Asia-Pac. J. Clin. Oncol.,2017
2. Cancer statistics, 2018;Siegel;CA Cancer J. Clin.,2018
3. Wang, S., Liu, Y., Feng, Y., Zhang, J., Swinnen, J., Li, Y., and Ni, Y. (2019). A Review on Curability of Cancers: More Efforts for Novel Therapeutic Options Are Needed. Cancers, 11.
4. Society, A.C. (2019). Breast Cancer Facts & Figures 2019–2020, American Cancer Society, Inc.
5. Computer-aided detection of breast cancer on mammograms: A swarm intelligence optimized wavelet neural network approach;Dheeba;J. Biomed. Inform.,2014
Cited by
13 articles.
订阅此论文施引文献
订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献