Model-based calibrations for medical 3D reconstructions

Obtaining high quality 3D reconstructions is of increasing importance to the medical society. For a successful mapping of 2D images, leading to a 3D reconstruction, highly accurate pose information of the systems imaging components, typically an X-ray source and detector, is required. A time consuming calibration is required to compensate for system imperfections, mechanical deflections and dynamical effects. In this work, a model-based approach is proposed to reduce the required calibration time while guaranteeing a desired reconstruction quality. Model parameters of the proposed model are identified using the currently used phantom-based calibration approach. A significant reduction in calibration times is obtained by exploiting the predictive properties of the model to predict variations between various scan types. Future work includes Observer-based filtering techniques to accurately predict dynamical behavior of the system.