Premium
SU‐E‐I‐03: Evaluation of Impacts of Two Assumptions in Cone Beam CT Geometry Calibration
Author(s) -
Xu Y,
Li H,
Yan H,
Cervino L,
Jiang S,
Jia X,
Zhou L
Publication year - 2013
Publication title -
medical physics
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.473
H-Index - 180
eISSN - 2473-4209
pISSN - 0094-2405
DOI - 10.1118/1.4814103
Subject(s) - ellipse , pixel , detector , calibration , geometry , rotation (mathematics) , cone beam computed tomography , optics , physics , trajectory , mathematics , range (aeronautics) , beam (structure) , computed tomography , statistics , medicine , materials science , astronomy , composite material , radiology
Purpose: Cone‐beam CT image quality crucially depends on the mechanical geometry accuracy. One of the state‐of‐art calibration methods utilizes two projected ellipses trajectory of two BBs moving along circular paths. The method, however, assumes that detector rotation angle about the axis in the source rotational plane is zero. It also ignores inter‐ellipse location constraints when more than two BBs are used and different ellipses are identified individually. The purpose of this study is to evaluate the necessity of these two issues. Methods: We simulated projections of our in‐house table CBCT system with realistic geometrical parameters. We particularly set the angle that were ignored in the ellipse‐based calibration method to be non‐zero. We have also simulated the cases with irregular precessions of the CBCT rotational axis, which make the projected BB trajectory not ellipses and hence demonstrate the effects of neglecting the inter‐ellipse correlations. For each case, the ellipse‐based geometry calibration is conducted. The calibrated pixel locations are compared with the ground truth. Results: The pixel size in our detector is 0.298 mm and source to detector distance is 1500mm. For detector rotation angles in the real machine range of 0∼1.0 degree, the maximum error of the calibrated pixel location is 0.596 mm (∼2 pixels) and the mean errors is up to 0.196 mm (∼0.7 pixels). For the cases with irregular precession angles of 0∼0.3 degree, the maximum and mean pixel location error is 0.894 mm (∼3 pixels) and up to 0.230 mm(∼0.8 pixels), respectively. Conclusion: Considering the inherent accuracy of the calibration method is 4.31 pixels (2.187mm), it is found that assuming a zero detector rotation angle and neglecting the inter‐ellipse constraints do not lead to serious degradations of calibration accuracy. This project is supported by Guangdong Strategic emerging industry core technology research (Grant No.2011A081402003) and Guangzhou Municipal special major science and technology programs (Grant No.11BppZLjj2120029).