Computer-aided characterization of pulmonary nodules based on 4-D curvatures and its usefulness for benign and malignant classification

摘要

This paper presents an approach to characterize pulmonary nodules on Multi-Slice CT images using four-dimensional (4-D) curvature indexes. In this approach, the 4-D principal curvatures are directly computed by the differential values of the 3-D CT images. In order to measure the intensity structure inside the nodule, two shape indexes are devised from the 4-D principal curvatures. Applied the shape indexes to benign and malignant nodules, we present the potential usefulness of the proposed measures to discriminate between benign and malignant nodules.