Proceedings of the IEEE International Conference on Biomedical Robotics and Biomechatronics (BioRob), pp. 383–388
June 2016 · Singapore, Singapore · doi: 10.1109/BIOROB.2016.7523656
Applications like cone beam computed tomographies (CBCTs) or 2D-3D overlays need a high geometrical accuracy of the C-arm system. In order to achieve this, geometry calibrations are performed to increase the accuracy given by the kinematics of the system. Commonly X-ray images of a phantom with known geometry are taken for the calibration. The images, together with a 3D model of the phantom, serve as input for an optimizer, which estimates the pose of the C-arm relative to the phantom. Afterwards, the estimates are used to increase the geometrical accuracy of the system. Inaccuracies due to real world effects appear, e.g. manufacturing or assembly inaccuracies of the phantom, the position of the X-ray tube, or the pose of the detector. To evaluate these factors separately a simulation is helpful, which needs to be as realistic as possible. To achieve this we defined three requirements, which have to be fulfilled: realistic noise, realistic absolute errors and similar error distributions within the working volume. By means of these criteria we investigate if our simulation mirrors a real world C-arm pose measurement for C-arm geometry calibration sufficiently.