MGRG-morphological gradient based 3D region growing algorithm for airway tree segmentation in image guided intervention therapy

Abstract

Accurate surgical planning and guidance plays an important role in successful implementation of image guided intervention. In interventional lung cancer diagnosis and treatments, precise segmentation of airway trees from lung CT images provides crucial visualization for preoperative planning and intraoperative guidance to avoid major trachea injury. While 3D region growing can segment main the parts of an airway tree (trachea, left and right main bronchus, as well as bronchi), the method fails at bronchiole segmentation and is not robust. Mathematical morphology is an anatomical detective. In this paper, we propose a morphological gradient based region growing (MGRG) algorithm to overcome the intensity inhomogeneity, and improve the robustness of 3D region growing on extraction of bronchioles. The MGRG algorithm is validated using lung CT images, and results show that it is able to segment bronchioles, and outperforms the traditional region growing method on airway tree segmentation.