Title: Cavitary nodule segmentation in computed tomography images based on self-generating neural networks and particle swarm optimisation

Authors: Juan-juan Zhao; Guo-hua Ji; Yong Xia; Xiao-long Zhang

Addresses: College of Computer Science and Technology, Taiyuan University of Technology, Shanxi, 030024, China ' College of Computer Science and Technology, Taiyuan University of Technology, Shanxi, 030024, China ' School of Computer Science, Northwestern Polytechnical University, Shaanxi, 710072, China ' College of Information Sciences and Technology, Pennsylvania State University, University Park, Pennsylvania, 16802, USA

Abstract: Lung nodule segmentation is an important pre-processing step for analysis of solitary pulmonary nodules in computed tomography (CT) imaging. However, the previous nodule segmentation methods cannot segment the cavitary nodules entirely. To address this problem, an automated segmentation method based on self-generating neural networks and particle swarm optimisation (PSO) is proposed to ensure the integrity of cavitary nodule segmentation. Our segmentation method first roughly segments the image using a general region-growing method. Thereafter, the PSO-self-generating neural forest (SGNF)-based classification algorithm is used to cluster regions. Finally, grey and geometric features are utilised to identify the nodular region. Experimental results show that our method can achieve an average pixel overlap ratio of 88.9% compared with manual segmentation results. Moreover, compared with existing methods, this algorithm has higher segmentation precision and accuracy for cavitary nodules.

Keywords: particle swarm optimisation; PSO; self-generating neural networks; clustering; nodule segmentation; cavitary nodules; computed tomography; CT images; medical imaging; lung nodules; pre-processing; pulmonary nodules.

DOI: 10.1504/IJBIC.2015.067999

International Journal of Bio-Inspired Computation, 2015 Vol.7 No.1, pp.62 - 67

Received: 03 Dec 2014
Accepted: 15 Dec 2014
Published online: 12 Mar 2015 *

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