Automatic lung surface registration using selective distance measure in temporal CT scans

Abstract

In this paper, we propose a novel technique of lung surface registration for investigating temporal changes such as growth rates of pulmonary nodules. For the registration of a pair of CT scans, a proper geometrical transformation is found through the following steps: First, optimal cube registration is performed for the initial gross registration. Second, for allowing fast and robust convergence on the optimal value, a 3D distance map is generated by the local distance propagation. Third, the distance measure between surface boundary points is repeatedly evaluated by the selective distance measure. Experimental results show that the performance of our registration method is very promising compared with conventional methods in the aspects of its computation time and robustness. © Springer-Verlag 2004.