Evaluation of deformable registration of patient lung 4DCT with subanatomical region segmentations Ziji Wu a Department of Radiation Oncology, Massachusetts General Hospital, 55 Fruit Street, Boston, Massachusetts 02114 Eike Rietzel Siemens Medical Solutions, Particle Therapy, Henkestrasse 127, 91052 Erlangen, Germany Vlad Boldea LIRIS Laboratory, Université Lumière Lyon, Lyon, France David Sarrut Léon Bérard Anti-Cancer Center, 28 rue Laënnec, 69373 Lyon, France and CREATIS Laboratory, UMR CNRS 5220, Inserm U 630 Lyon, France Gregory C. Sharp Department of Radiation Oncology, Massachusetts General Hospital, 55 Fruit Street, Boston, Massachusetts 02114 Received 10 May 2007; revised 28 November 2007; accepted for publication 30 November 2007; published 30 January 2008 Deformable registration is needed for a variety of tasks in establishing the voxel correspondence between respiratory phases. Most registration algorithms assume or imply that the deformation field is smooth and continuous everywhere. However, the lungs are contained within closed invaginated sacs called pleurae and are allowed to slide almost independently along the chest wall. This sliding motion is characterized by a discontinuous vector field, which cannot be generated using standard deformable registration methods. The authors have developed a registration method that can create discontinuous vector fields at the boundaries of anatomical subregions. Registration is performed independently on each subregion, with a boundary-matching penalty used to prevent gaps. This method was implemented and tested using both the B-spline and Demons registration algorithms in the Insight Segmentation and Registration Toolkit. The authors have validated this method on four patient 4DCT data sets for registration of the end-inhalation and end-exhalation volumes. Multiple experts identified homologous points in the lungs and along the ribs in the two respiratory phases. Statistical analyses of the mismatch of the homologous points before and after registration demon- strated improved overall accuracy for both algorithms. © 2008 American Association of Physicists in Medicine. DOI: 10.1118/1.2828378 Key words: deformable image registration, validation, 4DCT I. INTRODUCTION Deformable registration is used for a variety of tasks in four- dimensional 4D radiotherapy, including contour propaga- tion, treatment adaptation, dosimetric evaluation, and 4D optimization. 1–11 Several different core algorithms have been proposed and validated for nonrigid registration of CT im- ages for cancer patients. For example, Wang et al. used an accelerated Demons algorithm and evaluated it on prostate, head-and-neck, and lung cases. 12 They accelerated the De- mons algorithm by introducing an active force along with an adaptive force strength adjustment during the iterative pro- cess. The improvements led to not only a speedup over the original algorithm but also a high tolerance of large organ deformations. Yang et al. employed an in-house B-spline BSP image registration software with normalized cross- correlation metric to evaluate cone beam CT for dose calculation. 13 A variety of other methods include, but are not limited to, optical flow, thin-plate spline, calculus of varia- tions, and finite element methods with different motion models. 2,14–19 In nearly every implementation of deformable registra- tion, regularization and smoothness penalties are imposed to achieve a smooth and continuous deformation vector field. While regularization is required because deformable registra- tion is an ill-posed problem, it is difficult for these algo- rithms to create true discontinuities. In particular, these algo- rithms tend to have reduced accuracy near the pleural boundary, where the lungs can slide against the chest wall to create discontinuities of more than two centimeters. To solve this problem, we have developed an approach for registration of 4DCT that respects the discontinuity at the pleural inter- face. As described in Rietzel and Chen, 20 the thorax is seg- mented into moving lungs, mediastinum, and abdomen and less-moving the rest subregions, and each region is regis- tered separately. In this article, we introduce a boundary- matching criterion that helps to eliminate gaps between sepa- 775 775 Med. Phys. 35 „2…, February 2008 0094-2405/2008/35„2…/775/7/$23.00 © 2008 Am. Assoc. Phys. Med.