Detection algorithm of pulmonary nodules based on deep learning

Lung cancer has become one of the most harmful cancers to human health as air quality deteriorates worldwide and the number of smokers increases. The early stage of lung cancer is usually characterized by pulmonary nodules. If pulmonary nodules and their cancerization can be detected in time and treated early, it is of great value to improve the survival rate of patients with early lung cancer. In view of the problems of low accuracy, large detection range, low sensitivity and large number of false positives in the current detection model of pulmonary nodules, this paper proposed a detection algorithm of pulmonary nodules based on Faster-RCNN and U-net structure with residual module. The algorithm introduces a Convolutional Block Attention Module (CBAM), which combines channel and space, to realize adaptive feature learning and feature weight, and combines the idea of feature fusion to make full use of deep and shallow information to enhance feature extraction ability. It can improve the sensitivity of pulmonary nodules detection and assist radiologists to diagnose pulmonary nodules effectively.