Curvature based analysis of internal structure of pulmonary nodules using thin-section CT images

A curvature based representation is introduced for characterization, of benign and malignant pulmonary nodules in thin-section CT images. Through curvatures a local description of the pulmonary nodule images is obtained using shape index, curvedness, and CT value. Based on this local description, a multiscale shape spectrum is derived to globally represent the 3D pulmonary nodule image. The classification effectiveness of shape spectral features is evaluated for four different regions: (i) entire 3D pulmonary nodule; (ii) core region in the 3D pulmonary nodule; (iii) complement of the core region in the 3D pulmonary nodule; (iv) neighborhood region surrounding the 3D pulmonary nodule.