Research On Classification Algorithm Of Non-small Cell Lung Cancer Based On Pathological Image

In recent years,there are more and more cases of human cancer,especially lung cancer,which is a common malignant tumor.Most of them are non-small cell lung cancer（NSCLC）,while the sub subtypes of non-small cell lung cancer are mainly squamous cell carcinoma of the lung（LUSC）and lung adenocarcinoma（LUAD）.Due to the complex characteristics of pathological images,With the rapid development of computer vision,the use of deep learning and other technologies can effectively identify professional image categories and accurately distinguish their subtypes,which can help doctors diagnose the disease quickly and accurately to a great extent,and has very important clinical significance.Deep learning is a powerful tool in the field of computer vision.The integration of this technology and medical treatment provides a new idea for all aspects of lung cancer diagnosis,and promotes the development of artificial intelligence and the medical community.Aiming at the problems of complex texture and high similarity of pathological images of non-small cell lung cancer,which lead to low inspection efficiency and waste a lot of time,this paper designs a classification optimization algorithm based on deep learning,makes a detailed and in-depth study on the classification of pathological images of non-small cell lung cancer,and makes targeted improvements and optimization,in order to improve the recognition effect and accuracy of the model,The main contents of this paper are as follows:（1）A multi-attention-based dual-task classification algorithm for non-small cell lung cancer（Multi-Attention Dense Net,MADNet）is proposed.In view of the insufficient number of images in the data set and the uneven distribution of each type of images,this paper uses flip rotation and other methods to analyze the data.Enhancement to solve the problem of sample imbalance and insufficient number.The backbone network improves the original Densenet121 and integrates the improved attention dense connection module to simultaneously perform binary and multi-classification tasks on non-small cell lung cancer pathological images.The experimental results show that the improved network further improves the classification performance.（2）A lung cancer image classification model（SE-Res Net VGG Net,SE-RVNet）model based on two-stream transfer learning and attention mechanism is proposed.In the core module,VGG16 and Res Net50 are used for migration learning,the characteristic information output by the designated layer of VGG16 and Res Net50 is fused,and the initial weight of the network is changed.Then add the se attention mechanism module for feature extraction.FRN is introduced into the model for normalization,and focal loss function is used to balance the weight of positive and negative samples.The experimental results show a good recognition effect on TCGA data set,and the accuracy of two classification and three classification on TCGA data set is 99.8% and 95.9% respectively.In this paper,the deep learning algorithm is used to study the classification algorithm of pathological images of non-small cell lung cancer,and the existing classification algorithms are optimized and improved.Through many experimental comparisons,the algorithm proposed in this paper has obtained more efficient accuracy and good classification effect,which is also of great help to the development of medical field.