Study On The Detection Effect Of Pulmonary Nodules Based On Data Augmentation Method And Embedding Mechanism

Lung cancer is the most deadly cancer.The early manifestation of lung cancer is pulmonary nodules,so the early diagnosis of pulmonary nodules can be used to reduce the mortality of lung cancer.Computer aided diagnosis(CAD)based on deep learning can identify the suspected areas of pulmonary nodules in CT images,thus improving the accuracy and efficiency of CT diagnosis.In order to further develop and apply the deep learning-based pulmonary nodules assisted diagnosis technology,the accuracy and robustness of deep learning models for detection and classification of pulmonary nodules can be improved by increasing the data and improving the model structure.Due to the high cost of image labeling,it is difficult to provide a large number of data,while the generation model can automatically generate a large number of pulmonary nodules samples.The embedding mechanism can effectively improve the robustness of the model and improve the training effect.In this work,this paper explores the data augmentation method based on the generation model and the model structure improvement method based on the embedding mechanism.In terms of data augmentation,the main research contents of this paper are as follows: 1)proposing a new data augmentation strategy and introducing the embedding mechanism to improve the existing models.In the augmentation strategy,a 3D pixel-level statistics algorithm is proposed to generate pulmonary nodule samples.The result of the 3DVNet model with the augmentation strategy for pulmonary nodule detection shows that the proposed data augmentation method outperforms the method based on generative adversarial network(GAN)framework.2)The embedding mechanism is designed to better understand the meaning of pixels of pulmonary nodule samples by introducing hidden variables.The result of 3DVGG network with embedding mechanism for pulmonary nodule classification shows that the embedding mechanism improves the accuracy and robustness for the classification of pulmonary nodules obviously.