Room: Pins Room | Hall 2
Purpose: To intercompare two 3D/3D registration platforms, one 2D/3D registration method, and one 3D surface registration method. These three technologies are available to perform 6D registrations for image-guided radiotherapy treatment.
Methods: Twelve and eight fiducial markers were asymmetrically placed on the surface of a head phantom and a body phantom, respectively. The solution of the rotational transformation between the two image sets (planning CT and CBCT) was optimized through least-squares fitting of the fiducial positions using singular value decomposition (SVD). This was used as the gold standard to evaluate the rotational accuracy of 3D/3D registration in Varianâ€™s platform (3D3D_V), 3D/3D and 2D/3D registration in the BrainLab ExacTrac system (3D3D_E and 2D3D), as well as 3D surface registration in the AlignRT system (3DS). The phantoms were setup on a baseplate with random rotations in three axes (Rotation, Roll, and Pitch) within three degrees. 24 measurements (16 in head phantom, 8 in body phantom) of rotational shifts were made at five different intracranial isocenters using the SVD, the 3D/3D, 2D/3D, and 3D surface registration methods. Wilcoxon signed-rank test was utilized to compare the difference of each registration method with SVD. A p-value<0.05 was considered significant.
Results: Using SVD method as standard, all methods were significantly different at the Roll direction; 2D3D & 3DS were significantly different at Rotation direction; 3DS was significantly different at Pitch direction. Among the three axes, Roll direction was less well determined by all methods. Considering all rotational axis directions together, 3D3D_E and 3DS were significantly different. In terms of absolute rotational error at three directions (Rotation, Roll, Pitch), the meanÂ±SD values for 3D3D_V, 3D3D_E, 2D3D, and 3DS were (0.1Â°Â±0.1Â°,0.2Â°Â±0.2Â°,0.1Â°Â±0.1Â°), (0.1Â°Â±0.1Â°,0.3Â°Â±0.2Â°,0.2Â°Â±0.1Â°), (0.3Â°Â±0.1Â°,0.6Â°Â±0.2Â°,0.2Â°Â±0.1Â°), and (0.3Â°Â±0.1Â°,0.3Â°Â±0.3Â°,0.4Â°Â±0.3Â°), respectively.
Conclusion: A truly 3D3D registration showed insignificant rotational alignment error. Partial 3D registration may double the error. All methods are clinically acceptable.