Hybrid Attention Mechanism Guided Convolutional Neural Network for Breast Cancer Histology Images Classification
Yutong Zhong, Yan Piao, Guohui Zhang · 2021
Objective: Breast cancer is the most vulnerable type of cancer for women, examination of histopathological sections is the gold standard for diagnosing cancer. However, the high complexity and the increasing number of pathological images make the work of pathological diagnosis extremely time-consuming, and the sharp increase in the workload of pathologists will also cause the diagnosis results to be subjectively affected. Therefore, the development of an automated and accurate method for histopathological images analysis is critical to assist pathologists in their work. Methods: In this paper, we propose a deep learning method guided by the hybrid attention mechanism for fast and effective classification on the hematoxylin & eosin stained breast biopsy images. Firstly, this method takes advantages of DenseNet and utilizes the information of the feature map. Afterward, attention and channel attention are used to guide the extraction of the most useful visual features. Results: With the use of five-fold cross-validation, the best model obtaines an accuracy of 96.10% on the BACH2018 dataset. We also evaluate our method on other datasets, and the experimental results demonstrate that our model has reliable performance. Conclusions: This study indicates that our histopathological image classifier with a hybrid domain attention-guided deep-learning model for breast cancer shows significantly better results than the latest methods. It has great potential as an effective tool for automatic evaluation of digital histopathological microscopic images for computer-aided diagnosis.