Breast cancer detection using microscopic histopathological images: A transfer learning approach

The extraction of features from histological images is a challenging part of the computer-aided detection of breast cancer. This work presents a novel deep feature extraction engineering for subtypes of Breast detection-based transfer learning approach using the BreaKhis dataset. This approach consists of five phases: feature extraction, concatenation, transformation, selection, and classification. In the first phase, nineteen pre-trained convolutional neural networks were utilized as feature extractors to extract features from the input images. Support vector machines were used in this phase to compute the loss value function. In the classification phase, it was used as a classifier. According to the feature extraction results, the two networks attained the highest accuracies on the dataset, outperforming other networks. The two networks considered were chosen and concatenated to produce the DRNet model, a combination of ResNet50 and DenseNet201 pre-trained networks. The extracted features were decomposed into five sub-hand low-level features using a multilevel discrete wavelet transform in the transformation phase. Iterative neighborhood component analysis was utilized to choose the minimum number of features required in the classification phase. Cubic support vector machines and k-nearest neighbors were used as classifiers in the final phase. k-nearest neighbor produced an average classification rate accuracy of 98.74%, 97.7%, 97.3%, and 98% on the 40×, 100×, 200×, and 400× levels of magnification, respectively, where support vector machines produced an average classification rate accuracy of 98.61%, 98.04%, 97.68%, and 97.71% on the 40×, 100×, 200×, and 400× levels of magnification, respectively, using the one-gainst-all and 10-fold cross-validation method.