Verification of image registration for the commercial treatmetn planning systems

Abstract

Image registration is essential in treatment planning and position verification for highly radiation conformal delivery methods. The purpose of this study is to evaluate the accuracy of image registration in Varian Eclipse V. 8.6.17 and Oncentra V. 3.2.303 treatment planning system (TPS). Then the suitable image registration method of mutual information and match point between planning computed tomography and cone beam computed tomography are observed for 20 head and neck cases. The image registration verifications were performed in 2 TPSs using ImSim commercial QA software. The three set of images from ImSim commercial QA software, which the first one was stationary, the second set was 5, 10, 15 degree rotation and the third set was 5, 10, 15 mm translation in X, Y, and Z axes, were imported to Eclipse and Oncentra TPSs. The registration was made between the stationary and moving images. The registration errors were observed for mutual information and match point registration algorithms in each TPS. Then the planning CT image and cone beam CT image for 20 head and neck patients were registered on the Eclipse TPS by the mutual information and match point methods. The position shifts by the two methods were compared and evaluated with those registered optimization by radiation oncologist. For the registration of ImSim phantom, the Eclipse TPS shows slightly better registration than Oncentra TPS. The mutual information method demonstrates slightly less registration error than match point method, because pixel by pixel is used for mutual information while 4 points is selected for the match point registration method. The ImSim QA software can check both translation and rotation with no influence of setup position compared with the phantom or patient study. The registration error is increased when moving set is rotated >5 degree for all axes. For cone beam CT registration with planning CT of head and neck patients, the mean deviation from registered optimization by radiation oncologist for mutual information method are 2.02 ± 2.00 mm and 0.82 ± 1.16 degree and for match point method are 3.05 ± 2.92 mm and 1.31 ± 1.35 degree. The mutual information also shows superior image registration than match point method. From the result the mutual information and match point methods can assist the medical oncologist to optimize the image registration easily and reliably without image misalignment.