Lung segmentation method with dilated convolution based on VGG-16 network

Authors

Lei Geng, Tianjin Polytechnic University, Tianjin Key Laboratory of Optoelectronic Detection Technology and System
Siqi Zhang, Tianjin Polytechnic University, Tianjin Key Laboratory of Optoelectronic Detection Technology and System
Jun Tong, Tianjin Polytechnic University, Tianjin Key Laboratory of Optoelectronic Detection Technology and System, University of WollongongFollow
Zhitao Xiao, Tianjin Polytechnic University, Tianjin Key Laboratory of Optoelectronic Detection Technology and SystemFollow

RIS ID

138322

Publication Details

L. Geng, S. Zhang, J. Tong & Z. Xiao, "Lung segmentation method with dilated convolution based on VGG-16 network," Computer Assisted Surgery, vol. 24, (Sup2) pp. 27-33, 2019.

Abstract

Lung cancer has become one of the life-threatening killers. Lung disease need to be assisted by CT images taken doctor's diagnosis, and the segmented CT image of the lung parenchyma is the first step to help doctor diagnosis. For the problem of accurately segmenting the lung parenchyma, this paper proposes a segmentation method based on the combination of VGG-16 and dilated convolution. First of all, we use the first three parts of VGG-16 network structure to convolution and pooling the input image. Secondly, using multiple sets of dilated convolutions make the network has a large enough receptive field. Finally, the multi-scale convolution features are fused, and each pixel is predicted using MLP to segment the parenchymal region. Experimental results were produced over state of the art on 137 images which key metrics Dice similarity coefficient (DSC) is 0.9867. Experimental results show that this method can effectively segment the lung parenchymal area, and compared to other conventional methods better.