Nodule Detection in Lung CT Images Based on Deep Convolutional Neural Network

In order to increase the precision of detection and overcome the shortcomings of current methods, this paper presents a new approach to detect pulmonary CT images using deep convolutional neural networks. The model can extract image information of various sizes and features based on the multiscale convolution, which can improve the sensitivity of small nodules and low contrast nodules. Moreover, an adaptive weight adjusting mechanism is used to adjust the importance of each area, and the effect of background noise can be suppressed effectively, and the regional characteristic of the node can be enhanced. The performance of this model is evaluated on the publicly available LIDC-IDRI dataset. The experiment results indicate that the accuracy, recall rate and F1 score are increased <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$12 \%, 8 \%$</tex> and 9%, respectively. In particular, the model has obvious superiority in detecting small nodules. Moreover, this model has a good performance in reasoning and real time, so it can be used in clinic. The proposed deep convolutional neural network model is proved to be efficient and stable, and provides a new way to develop intelligent medical imaging diagnostic system.