A Classification Method of Breast Pathological Image Based on Residual Learning
Ma Xuru · 2020
Today, breast cancer has become a globally recognized tumor disease with high incidence and high mortality. The use of advanced convolutional neural network for the auxiliary diagnosis of breast pathological images is the general trend. Due to the complex background of the medical pathology image and more noise, a bilateral filter is used for noise reduction. Aiming at the problem of refinement, complexity and weakening of the morphological texture features of breast pathological images, the idea of using residual network residual units to simplify the learning process and enhance gradient propagation is adopted. The deep residual network is used to divide the image into benign and malignant tumors. For medical images with billions of pixels, the strategy of randomly extracting image patches is employed for data enhancement, and the model is evaluated using summation rules. Experimental results show that the recognition accuracy of the model reaches 96%.