Journal of Northeastern University Natural Science ›› 2018, Vol. 39 ›› Issue (11): 1556-1561.DOI: 10.12068/j.issn.1005-3026.2018.11.008

• Information & Control • Previous Articles     Next Articles

Generative Adversarial Networks for the Classification of Lung Nodules Malignant

XU Jiu-qiang, HONG Li-ping, ZHU Hong-bo, ZHAO Hai   

  1. School of Computer Science & Engineering, Northeastern University, Shenyang 110169, China.
  • Received:2017-07-31 Revised:2017-07-31 Online:2018-11-15 Published:2018-11-09
  • Contact: HONG Li-ping
  • About author:-
  • Supported by:
    -

Abstract: In order to solve the proportion of benign and malignant lung nodule, a novel model named deep convolutional generative adversarial networks(DCGAN) was introduced. The model generates lung nodule images with similar texture feature from the input lung nodules images, and then using them to train the DCGAN model. In addition, the classification of image source is changed to the classification of image source and lung nodules grade 1~5. thus, the noise immunity of DCGAN model is enhanced and the classification of lung nodules by DCGAN model is realized. Experiments show that the model G in improved DCGAN enhances the performance of anti-noise capability with 90.42% images are distinguished true images when it generating images, and the model D has a good discriminant ability for the classification of lung nodule images and the classification accuracy of lung nodules is 70.89%, the recognition rate of malignant lung nodules is 80.13%.

Key words: lung nodules, DCGAN (deep convolutional generative adversarial networks), texture feature, improved DCGAN, lung nodules rank classification

CLC Number: