An accuracy assessment of different rigid body image registration methods and robotic couch positional corrections using a novel phantom

Abstract

Image guided radiotherapy (IGRT) using cone beam computed tomography (CBCT) images greatly reduces interfractional patient positional uncertainties. An understanding of uncertainties in the IGRT process itself is essential to ensure appropriate use of this technology. The purpose of this study was to develop a phantom capable of assessing the accuracy of IGRT hardware and software including a 6 degrees of freedom patient positioning system and to investigate the accuracy of the Elekta XVI system in combination with the HexaPOD robotic treatment couch top. The constructed phantom enabled verification of the three automatic rigid body registrations (gray value, bone, seed) available in the Elekta XVI software and includes an adjustable mount that introduces known rotational offsets to the phantom from its reference position. Repeated positioning of the phantom was undertaken to assess phantom rotational accuracy. Using this phantom the accuracy of the XVI registration algorithms was assessed considering CBCT hardware factors and image resolution together with the residual error in the overall image guidance process when positional corrections were performed through the HexaPOD couch system. The phantom positioning was found to be within 0.04 (σ = 0.12)°, 0.02 (σ = 0.13)°, and -0.03 (σ = 0.06)° in X, Y, and Z directions, respectively, enabling assessment of IGRT with a 6 degrees of freedom patient positioning system. The gray value registration algorithm showed the least error in calculated offsets with maximum mean difference of -0.2(σ = 0.4) mm in translational and -0.1(σ = 0.1)° in rotational directions for all image resolutions. Bone and seed registration were found to be sensitive to CBCT image resolution. Seed registration was found to be most sensitive demonstrating a maximum mean error of -0.3(σ = 0.9) mm and -1.4(σ = 1.7)° in translational and rotational directions over low resolution images, and this is reduced to -0.1(σ = 0.2) mm and -0.1(σ = 0.79)° using high resolution images. The phantom, capable of rotating independently about three orthogonal axes was successfully used to assess the accuracy of an IGRT system considering 6 degrees of freedom. The overall residual error in the image guidance process of XVI in combination with the HexaPOD couch was demonstrated to be less than 0.3 mm and 0.3° in translational and rotational directions when using the gray value registration with high resolution CBCT images. However, the residual error, especially in rotational directions, may increase when the seed registration is used with low resolution images.