HEAD AND NECK LYMPH NODE REGION DELINEATION USING A HYBRID IMAGE REGISTRATION METHOD Chia-Chi Teng 1 , Linda G. Shapiro 1,2 , Ira Kalet 2,3 1 Department of Electrical Engineering, 2 Department of Computer Science 3 Department of Radiation Oncology, University of Washington ABSTRACT The success of radiation therapy depends critically on accurately delineating the target volume, which is the region of known or suspected disease in a patient. Methods that can compute a contour set defining a target volume on a set of patient’s biomedical images will contribute greatly to the success of radiation therapy and drastically reduce the workload of radiation oncologists, who currently often draw the targets by hand on images using simple computer drawing tools. We are developing methods for automatically selecting and adapting standardized regions of tumor spread based on the location of lymph node regions in a standard or reference case, using image registration techniques. Previously available image registration techniques (deformable transformations computed using mutual information [5]) appear promising and can be supplemented by utilizing landmark correspondences in the optimization process to come closer to achieving a clinically acceptable match. 1. INTRODUCTION With the rapid development of conformal radiation therapy and Intensity Modulated Radiation Therapy (IMRT) systems in the field of Radiation Oncology, it is now possible to deliver a precise dose of radiation to irregularly shaped target (tumor) volumes. A 3-dimensional target volume needs to be defined to facilitate a treatment plan, and the success of the treatment depends on knowing the exact extend of the target volume in each patient. Radiation oncologists have adopted definitions for various components of the target volume. The Gross Target Volume (GTV) is the visible and palpable tumor mass usually visible on images (CT and MR). It is not automatically identifiable with existing image processing techniques. The Clinical Target Volume (CTV) includes the locations of microscopic local and regional spread, which usually means the GTV plus the lymph node regions around it. Microscopic disease cannot currently be imaged by any existing clinical technique. Even the nodes themselves are often hard to identify in the images. The task of delineating these nodal regions, which is usually done by clinicians, is quite time consuming. Clinicians often elect to perform less aggressive, non-conforming treatment, because they do not have the time to draw the outlines of the nodal regions and CTV, even if they are confident of which node groups are likely to have disease to treat. Inter-subject image registration methods have been the subject of extensive study in many areas of biomedical imaging applications, for example the brain mapping. We are proposing an image registration method that maps predefined head and neck nodal regions of reference (canonical) models to target patients and suggests where the CTV might be in a target image set. 2. IMAGE BASED NODAL CLASSIFICATION Cervical lymph nodes are divided into regions or “levels” that are described by their anatomical locations. Traditional classification has used surgical landmarks or other physical assessment criteria, but more recently image-based classifications have been proposed [1][2] that provide a more consistently reproducible nodal staging model. Figure 1 shows an example of nodal region contours in a CT image. 3. LANDMARK CORRESPONDENCE Anatomical landmarks are commonly utilized in image registration methods, which use landmark points to match the image properties in different image sets and bring them into alignment. They are also often used in combination with an entirely different registration basis, such as brain mapping. For example, Christensen et al. [3] use interactive methods to locate anatomical landmarks that define the Frankfort Horizontal Plane, the Median Sagittal Plane, and the Coronal Plane; they then use those planes to estimate the rigid registration before running non-rigid registration with a different method (elastic model). We have chosen the mandible and hyoid bones as our landmarks because of 462 0-7803-9577-8/06/$20.00 ©2006 IEEE ISBI 2006