Potential usefulness of curvature based description for differential diagnosis of pulmonary nodules

Abstract

This paper examines the problem of obtaining a representation of the three-dimensional(3-D) pulmonary nodule images, which is a key problem in discriminating benign and malignant nodules for differential diagnosis of the lung cancer using thin-section CT images. A curvature based approach is developed with the aim of characterizing internal intensity structures of benign and malignant nodules. This approach makes use of curvature indexes to represent locally each voxel in a three-dimensional (3-D) pulmonary nodule image. From the distribution of curvature indexes and CT value over the 3-D pulmonary nodule image a set of histogram features is computed for global characterization of benign and malignant nodules. Linear discriminant analysis is used for classification and leave-one-out method is used to evaluate the classification accuracy. Compared with the performance of experienced physicians the potential usefulness of the curvature based features in the computer-aided differential diagnosis is demonstrated by using receiver operating characteristic (ROC) curves as the performance measure.