A new technique for multi-modal 3D image registration Gjenna Stippel 1 , Jim Ellsmere 2 , Simon K. Warfield 2 , William (Sandy) M. Wells 2 , and Wilfried Philips 1 1 Ghent University, Department TELIN, Gent, Belgium 2 Harvard Medical School, Brigham & Women’s Hospital, Boston, MA USA http://www.gjenna.com Abstract. In this paper we address the problem of multi-modal co- registration of medical 3D images. Several techniques for the rigid regis- tration of multi-modal images have been developed; in one of those the Kullback-Leibler distance is used to align 2D-3D angiographic images [1]. In this paper we investigate the performance of this technique on the registration of pairs of 3D CT/US images. We study the effects of various image perturbations on the performance of the registration, and obtain promising results. 1 Introduction Registering intraoperative ultrasound (US) to preoperative images like computed tomography (CT) or magnetic resonance (MR) remains a challenging problem in the field of image guided surgery. For many surgical applications, ultrasound provides adequate clinical information to carry out the required intervention. There are some applications, however, where being able to interpret the ultra- sound images in the context of the higher quality preoperative imaging has been shown to be helpful [2]. In this paper, we introduce a new rigid registration technique that aligns US and CT images by applying linear transformations to the coordinate system in order to minimize the Kullback-Leibler distance (KLD) between the observed joint intensity distribution, and a reference distribution representing the prior knowledge of the expected joint intensity histogram when the images are aligned properly [1]. We calibrate our KLD registration by taking an ultrasound scan, and man- ually aligning it with a CT scan to establish the fixed reference distribution. Afterwards, as more 3D ultrasound acquisitions are acquired over the course of the surgery, this calibration does not need to be repeated. Hence, the cost of the calibration is amortized over all the follow up scans during the procedure. If there are many scans, the calibration cost is a small percentage of the total effort. The performance of the method and its robustness under various image per- turbations is studied.