Search

Although computer-aided diagnosis (CAD) have shown excellent performance in Breast cancer (BC) histopathological image, it commonly requires a high-level network, and the recognition efficiency is frequently unsatisfied due to the complex structure of histopathological image. In this study, a ten-layers convolutional neural network (CNN) model called "ColorDeep" is used to extract color features corresponding to the different tissue parts in cell, pure color image slices obtained by a three-channel separation and reconstruction method are used as model input. Two models are tested on the BreaKHis dataset show that images under four magnifications achieved the recognition accuracy of 96.89%–99.67% at the image level, which is better than many state-of-the-art methods. The characteristics contained by the B channel have the largest effect on BC recognition, and compared to other research results, the proposed model improves the recognition speed on a single image by about 0.1s. More importantly, instead of using large histopathological images to input into the model for BC diagnosis, and instead of segmenting the nuclei, only the reconstructed B-channel features containing the nuclei region in the stained BC image need to be input into the model to enable accurate diagnosis of BC. • Three-channel reconstructed slices contain better color features of breast cancer. • Reconstructed B-channel features contain more information about the cell nucleus. • Using color features to diagnose breast cancer can avoid segmentation of nuclei. • Reconstructed color slices enable low-complexity model to identify breast cancer. • Considering recognition accuracy and speed, feature fusion model is more suitable for breast cancer diagnosis. [ABSTRACT FROM AUTHOR]

Breast cancer is regarded as the leading killer of women today. The early diagnosis and treatment of breast cancer is the key to improving the survival rate of patients. A method of breast cancer histopathological images recognition based on deep semantic features and gray level co-occurrence matrix (GLCM) features is proposed in this paper. Taking the pre-trained DenseNet201 as the basic model, part of the convolutional layer features of the last dense block are extracted as the deep semantic features, which are then fused with the three-channel GLCM features, and the support vector machine (SVM) is used for classification. For the BreaKHis dataset, we explore the classification problems of magnification specific binary (MSB) classification and magnification independent binary (MIB) classification, and compared the performance with the seven baseline models of AlexNet, VGG16, ResNet50, GoogLeNet, DenseNet201, SqueezeNet and Inception-ResNet-V2. The experimental results show that the method proposed in this paper performs better than the pre-trained baseline models in MSB and MIB classification problems. The highest image-level recognition accuracy of 40×, 100×, 200×, 400× is 96.75%, 95.21%, 96.57%, and 93.15%, respectively. And the highest patient-level recognition accuracy of the four magnifications is 96.33%, 95.26%, 96.09%, and 92.99%, respectively. The image-level and patient-level recognition accuracy for MIB classification is 95.56% and 95.54%, respectively. In addition, the recognition accuracy of the method in this paper is comparable to some state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Pathological examination is the gold standard for breast cancer diagnosis. The recognition of histopathological images of breast cancer has attracted a lot of attention in the field of medical image processing. In this paper, on the base of the Bioimaging 2015 dataset, a two-stage nuclei segmentation strategy, that is, a method of watershed segmentation based on histopathological images after stain separation, is proposed to make the dataset recognized to be the carcinoma and non-carcinoma recognition. Firstly, stain separation is performed on breast cancer histopathological images. Then the marker-based watershed segmentation method is used for images obtained from stain separation to achieve the nuclei segmentation target. Next, the completed local binary pattern is used to extract texture features from the nuclei regions (images after nuclei segmentation), and color features were extracted by using the color auto-correlation method on the stain-separated images. Finally, the two kinds of features were fused and the support vector machine was used for carcinoma and non-carcinoma recognition. The experimental results show that the two-stage nuclei segmentation strategy proposed in this paper has significant advantages in the recognition of carcinoma and non-carcinoma on breast cancer histopathological images, and the recognition accuracy arrives at 91.67%. The proposed method is also applied to the ICIAR 2018 dataset to realize the automatic recognition of carcinoma and non-carcinoma, and the recognition accuracy arrives at 92.50%. [ABSTRACT FROM AUTHOR]