الفهرس 
A transfer-learning-based framework for multi-class classification of breast cancer using whole-slide images
SummaryOnly 14 pages are availabe for public view
 |  | from | 94 |  |  |
| | | from | 94 | | |
Abstract
Breast cancer has become one of the most common death causes worldwide, especially in females. The diagnosis of breast cancer with haemo toxylin and eosin (H&E) stained slides is essential but non-trivial, and pathologists often differ on the final decision. Computer-aid diagnosis using whole-slide images (WSIs) helps to reduce the cost and improve the accuracy of the diagnosis. A transfer learning framework for the classification of H&E-stained breast biopsy images is proposed .The imaged tissues are divided into four classes, normal, benign lesion, in-situ carcinoma, and invasive carcinoma. Feature extraction in the proposed framework is based on DenseNet-201pre-trained convolution neural network (CNN) model. Also, the final classification layers are appropriately modified. Training and testing of the classification framework was carried out on the Bre Ast Cancer Histology (BACH) WSI datasets of the International Conference on Image Analysis and Recognition (ICIAR) 2018 challenge. Image augmentation techniques were applied to increase the training data samples.The proposed framework achieved an average testing accuracy of about 95%