Research of Image Segmentation on Pulmonary Nodules in PET-CT Image Based on LIBSVM
QIANG Yan -, Yue Li · International Journal of Digital Content Technology and its Applications · 2013
Nowadays，it's still had to segment the Pulmonary nodules with others in PET-CT image. In this paper, to the Pulmonary nodules in PET-CT image, we use the method of LIBSVM to extract. As SVM having unique advantage in the problem of nonlinear identification, and there are more default parameters and less parameters adjustment in SVM when using LIBSVM, it's rapid and accurate to use LIBSVM to extract the Pulmonary nodules in PET-CT image. In the experiment of this paper, we use RGB values and grayscale statistics of eigenvalues as different set of characteristics to extract the same Pulmonary nodules in PET-CT image. The experimental results show that, in the cost of time and accuracy of classification, LIBSVM using grayscale statistical characteristics is more accurate, and the performance of classification is better.