Research on CT Pulmonary Nodule Detection Based on Residual Module
CT is the primary imaging tool for obtaining lung images.In view of the low sensitivity of CT detection of lung nod-ules,this paper proposes a network based on residual module for the detection of lung nodules.Firstly,ResNet is used to replace the backbone of the U-shaped network,and the residual blocks are flexibly added.The Shortcut Connections are used to better connect the feature information of the context,and at the same time,the problems of network degradation and gradient disappearance are re-duced partly.In addition,the region proposals generated by RPN are directly used as output results in the output layer of the net-work,and Soft-NMS is used to suppress the redundant frames,which greatly preserves the effective characteristics of the nodules.The proposed algorithm is verified in the LUNA16 datasets.The CPM value reaches 0.887,which is 4.9%,1.7%and 4.8%higher than those of the other three models respectively.The sensitivity of pulmonary nodule detection is improved to a certain extent,which provides a theoretical reference for clinical examination.
万方数据
维普
