Quantification of shrinkage of lung lobe from chest CT images using the 3D extended Voronoi division and its application to the benign/malignant discrimination of tumor shadows

Abstract

In this paper, we propose a method to quantify the convergence of tissue in lung caused by cancer, using the three-dimensional (3D) extended Voronoi division from chest X-ray CT images. The convergence is the phenomenon that the pulmonary tumors pull tissues such as blood vessels and interstitium toward the pulmonary tumors. Because the malignant tumors (especially adenocarcinoma) often accompanies the convergence, quantification of the convergence is useful for the benign/malignant discrimination. We first calculate the Voronoi diagrams using the vessel regions which were classified according to the radius of vessels. Then a feature was calculated using the volumes of the Voronoi regions in the whole of lung region and those of the neighborhood of tumor region. The significant correlation between the proposed feature and the existence of the convergence is shown experimentally using actual CT images.