SU-G-206-04: A Method for Realizing Phantom Calibration and Geometry Calibration Accurately Based On a Geometry Evaluation Index




For traditional geometric calibration, the calibration accuracy relies on both accuracy of geometry phantom manufacture and accuracy of ball bearings (BB) location estimation. In this work, we have developed a method to perform phantom calibration and geometry calibration iteratively and accurately in a whole procedure.


We have designed and manufactured a geometry phantom with BB and an evaluation phantom of a crystal ball contained in a cubic gel box. Our calibration method consists of five steps: 1) Estimate BB locations using spiral CT image, which are then used to initialize the particles in Particle Swarm Optimization (PSO) algorithm; 2) Perform geometric calibration; 3) Reconstruct the images of the evaluation phantom based on the current geometry calibration; 4) Evaluate the reconstructed images using a geometry evaluate index; 5) Update BB locations in PSO algorithm. Repeat step2)-5) until our stopping criteria is met. The edge of the crystal ball in the calibration phantom on CBCT images is detected by Hough transform to define two circular rings outside and inside the ball. The evaluation index used in step 4) is defined as the difference of the averaged image intensities of these two circular rings.


We have demonstrated the feasibility and performance of our method on a benchtop CBCT system. It is observed that inaccurate BB locations lead to severe image distortion and relative small evaluation index. With our method, streak artifacts are reduced and the structure becomes sharper and clearer. The evaluation index increases fast within 10 iterations, and then becomes stable gradually.


Our method can perform accurate phantom calibration and geometry calibration together in a whole procedure. It helps to mitigate the impact of the geometry phantom manufacture errors on the calibration, which could hence save the cost of the geometry phantom manufacture.