Multi-View contourless registration of bone structures using a single calibrated flouroscope

1. OVERVIEW We present an extended approach for the accurate simultaneously registration of several X-ray images with a corresponding computer-tomography dataset (CT). The X-ray images are taken one after another using only a single uoroscope. The main idea of our approach is an iterative optimization, that incorporates a least-squares-criterion for evaluating the pixel-based diierence between each X-ray image and a corresponding synthetic image. Such a synthetic image is determined by projecting a part of the CT-dataset with respect to a given pose. The main beneets of our approach are the following: (1) no need for time-consuming intraoperative segmentation of X-ray images, (2) high robustness due to our multiview approach using several images with appropriate angle diierence, (3) high accuracy due to exact X-ray uoroscope calibration incorporating exterior and interior camera modelling as well as image dewarping. Our approach appears suitable for a variety of clinical applications. We have considered three of them in more detail: (1) navigation of a surgical instrument intraoperatively, (2) intraoperative localization for the insertion of pedicle screws, (3) robot-assisted total knee replacement. The presented registration method is based on a direct comparison of CT and intraop-erative X-rays. Therefore it is essential to avoid the appearance of artiicial landmarks, which are normally used intraoperatively for the calibration of an X-ray camera. Hence.