Deep Learning And Transfer Learning Based Breast Cancer Image Classification

The development of medical imaging technology and computer vision has provided powerful help for the diagnosis of breast cancer.In medical image classification,the complexity and diversity of breast cancer pathological tissues bring great obstacles to the accurate classification of pathological images.How to effectively use the characteristic information of pathological tissue images to provide reliable basis for the classification of tumor tissue pathological images has become a research hotspot.On the other hand,due to the particularity of medical images,there are relatively few large-scale and accurately labeled medical image data,leading to the problem of data dependence in medical image classification.In view of the above challenges,this paper proposes a breast cancer pathological image classification method based on deep learning and transfer learning.Based on deep learning and transfer learning theory,attention mechanism and convolutional neural network were introduced to improve the network model,and a transfer learning method based on optimal fine-tuning algorithm was proposed.Through the combination of the two methods,a breast cancer histopathological image classification method based on deep learning and transfer learning is implemented.The experimental results show that this method can greatly improve the accuracy of breast cancer histopathological image classification.The main research work is as follows:（1）To solve the problem of staining difference in Break His histopathological image data set,the staining standardization method was adopted to solve the problem of staining difference without destroying image structure information;Data enhancement method is used to solve the problem of insufficient image quantity and uneven distribution in original data set.（2）Based on Inception V3 network,an attention mechanism SE module is introduced to make full use of the information between feature channels to achieve an attention mechanism-based breast cancer histopathological image classification network.Compared with Alex Net and unimproved Inception V3,experimental results show that the classification accuracy of the improved network model is higher than that of the comparative network.（3）Transfer learning strategies under different conditions are discussed,and an optimal fine-tuning algorithm based on improved network model is proposed.The algorithm uses iterative competition to find the optimal fine-tuning depth,and the feasibility of the method is verified by experiments.Finally,breast cancer histopathological image classification methods based on deep learning and transfer learning are compared with Alex Net and Inception V3 models.The results show that the proposed method significantly improves the performance of network classification,and the classification accuracy is better than the comparison model.In the binary classification experiment of Break His data set,the classification accuracy reached88.5%.