Title: An automatic detection model of pulmonary nodules based on deep belief network

Authors: Zhiyong Zhang; Jialing Yang; Juanjuan Zhao

Addresses: Shanxi Province 109 Hospital, Taiyuan 030006, China; College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China ' Shanxi Province 109 Hospital, Taiyuan 030006, China; College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China ' Shanxi Province 109 Hospital, Taiyuan 030006, China; College of Computer Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China

Abstract: Deep belief network (DBN) is a typical representative of deep learning, which has been widely used in speech recognition, image recognition and text information retrieval. Owing to a large number of CT images formed by the advanced spiral CT scanning technology, a pulmonary nodules detection model based on user-defined deep belief network with five layers (PndDBN-5) is proposed in this paper. The process of the method consists of three main stages: image pre-processing, training of PndDBN-5, testing of PndDBN-5. First, the segmentation of lung parenchyma is done. Segmented images are cut with minimum external rectangle and resized using the bilinear interpolation method. Then the model PndDBN-5 is built and trained with pre-processed training samples. Finally, testing PndDBN-5 with pre-processed testing samples is completed. The data used in this method are derived from The Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) which is the largest open lung nodule database in the world. The experimental results show that the correct rate of PndDBN-5 model for pulmonary nodule detection reached 97.5%, which is significantly higher than the traditional detection method.

Keywords: deep belief network; contrastive divergence; pulmonary nodules; detection; CT.

DOI: 10.1504/IJWMC.2019.097415

International Journal of Wireless and Mobile Computing, 2019 Vol.16 No.1, pp.7 - 13

Available online: 16 Jan 2019 *

Full-text access for editors Access for subscribers Purchase this article Comment on this article