Journal of Northeastern University Natural Science ›› 2018, Vol. 39 ›› Issue (2): 181-185.DOI: 10.12068/j.issn.1005-3026.2018.02.007

• Information & Control • Previous Articles     Next Articles

Pulmonary Nodules Detection Based on 3D Features from CT Images

WANG Bin, ZHAO Hai, ZHU Hong-bo, PAK Chun-hyok   

  1. School of Computer Science & Engineering, Northeastern University, Shenyang 110169, China.
  • Received:2016-08-10 Revised:2016-08-10 Online:2018-02-15 Published:2018-02-09
  • Contact: WANG Bin
  • About author:-
  • Supported by:
    -

Abstract: A detection method based on the center continuity is proposed to improve the performance of pulmonary nodules detection. CT images are segmented by using SLIC (simple linear iterative clustering) superpixel algorithm in the method. Superpixels are merged according to the similarity to extract pulmonary areas and suspected pulmonary nodule areas, which reduces the missing rate of suspected pulmonary nodules. Suspected pulmonary nodules are diagnosed as positive which keeping center continuous in 3D space. All of CT images in experiments are obtained from Shanghai chest hospital and LIDC database. The experimental results of the improved algorithm show that sensitivity is 86.36% and false positive is 1.76.

Key words: pulmonary nodule, pulmonary nodule detection, pulmonary nodule segmentation, superpixel, SLIC (simple linear iterative clustering), center continuity

CLC Number: