Research on the classification of breast pathological images based on fusion of deep network models
The pathological analysis of breast tissue biopsy has important clinical application value.Aiming at the problems of time-consuming,labor-intensive,and incomplete extracted features in the manual extraction feature classification algorithm,this study combines with deep learning and proposes a model fusion method based on multi-stage migration and attention mechanism for benign and malignant breast pathological images classification.In order to speed up the training convergence speed and use the image features of different pathological image datasets,this paper adopts multi-stage transfer learning,and at the same time adds an attention mechanism to the network,and suppresses unnecessary features by learning important features in image channels and space classification accuracy.Finally,in order to utilize the features of images of different multiples of the dataset at the same time,a model fusion network is established for classification.The network achieves an AUC of 0.946 for classifying benign and malignant images.The experimental results show that the model fusion method based on multi-stage transfer and attention mechanism has achieved high accuracy in the classification of breast pathological images,which has positive guiding significance for breast cancer diagnosis.
breast cancerpathological image classificationmulti-stage transferattention mechanismmodel fusion
NETL
NSTL
万方数据
