Breast Cancer Classification from Histological Images using Multi-scale Dense Network
Yiping Zhou, Can Zhang, Shaoshuai Gao · 2022 2nd International Conference on Consumer Electronics and Computer Engineering (ICCECE) · 2022
Early diagnosis and treatment of breast cancer is essential to increase the survival rate. Currently, histological analysis performed by pathologists is critical to breast cancer diagnosis. Compared with pathologists, computer-aided system utilizes machine learning techniques can reduce workload and increase the classification accuracy. In this paper, a method based on the multi-scale dense network (MSDNet) was proposed for breast cancer histological image classification. Networks with variable depths were used to increase computational efficiency on the premise of prediction accuracy. The BACH 2018 dataset were used and classified into four classes: normal tissue, benign lesion, in situ carcinoma and invasive carcinoma. Experimental results show that the overall accuracy is 98.28% and 97.31 % for binary and multi-class classification in patch level, respectively. The image-wise classification accuracy reaches 100% by majority vote fusion decision after patch prediction. Meanwhile, the computational time is reduced almost 50% compared with other methods.