Abstract

Info

Title:: Support vector machine-based algorithm for the extraction and recognition of ground glass nodules in lung CT image

Author(s):: XU Ya’nan1; ZHAO Wei2; LI Ming2; SHI Hongli1; 1. School of Biomedical Engineering, Capital Medical University, Beijing 100069, China; 2. CT Room, Huadong Hospital Affiliated to Fudan University, Shanghai 200040, China

Keywords:: ; Keywords: lung; ground glass nodule; support vector machine; computed tomography; three-dimensional image

Abstract:: Abstract: A new approach based on support vector machine is proposed to extract and recognize the ground glass nodules in three-dimensional (3D) lung computed tomography (CT) image. Firstly, the lung parenchyma regions are segmented according to the 3D connectivity of the lung parenchyma. Then the isolated tissues which may be ground glass nodules are extracted from the lung parenchyma region. After the 3D shape features and 3D texture features are calculated, a linear model is established using these features to recognize the ground glass nodules. The coefficients of the model are determined by support vector machine based on the CT images labeled by clinicians. Finally, the linear model is used to recognize ground glass nodules from the isolated tissues. Among the labeled lung CT images of 139 patients in this study, the images of 100 patients were used as training set and the others as test set. The test results show that the proposed approach can effectively recognize the ground glass nodules in lung CT image. The area under receiver operating characteristic curve reaches 0.937 2.