A One-Stage Lung Nodule Detection Method Based on 3D Inception and Pyramid Fusion

Abstract

Lung cancer is a colossal danger to human life. Early detection of lung nodules has become the main means to deal with lung cancer. This paper proposes a one-stage lung nodule detection method based on 3D Inception and pyramid fusion. First, a 3D Inception module is designed to improve the adaptability of the network to multi scales and optimize semantic features. Second, the extracted features are input into the pyramid fusion module to obtain the deep semantic information and shallow position information at the same time. In addition, the Focal loss function is introduced to alleviate the imbalance between the samples in the training. Finally, experiments are carried out on the LUNA16 data set. Compared with other methods, the method in this paper obtains better detection results.