Abstract: In a Computer-Aided Detection (CAD) scheme for pulmonary nodules using computed tomography (CT) images, nodule segmentation is an important intermediate step, which impacts a great influence on the final performance of detection. In order to improve the detection rate of nodule and suppress the false positive, a more effective and physical meaningful nodule segmentation method is proposed in this paper. The algorithm is based on mean shift clustering method and CI (Convergence Index) features, which could represent the multiple Gaussian model of pulmonary nodules both for solid and sub-solid, substantially. This approach is based on an idea of utilizing features in a more "active" way, that is, we integrate the feature to the segmentation algorithm rather than just calculate them after segmentation. The presented segmentation method can figure out the outline of pulmonary nodules more precisely and especially suitable for the segmentation of sub-solid nodules.

Key words: nodule segmentation, solid nodule, sub-solid nodule, CI feature, mean shift algorithm, CT images, nodule segmentation, solid nodule, sub-solid nodule, CI feature, mean shift algorithm