Interventional 3D-angiography: calibration, reconstruction and visualization system

3D reconstruction of arterial vessels from planar radiographs obtained form multitude of angles around the object has gained increasing interest. The motivating application has been interventional angiography. In these procedures data acquisition, visualization, decision making, and the action are all incorporated in one loop that is time critical. This project encompasses an entire workflow; from system calibration, data acquisition, and image reconstruction, to the effective presentation of the results for the right action. Neurointerventios are carried out on X-ray angiography systems similar to the one illustrated in Figure 1. In this figure the smaller C-arm rotates about the patient’s head by roughly ° 200 . A total of 50-100 projections are acquired that are transferred to a post-processing workstation for 3D reconstruction. The 3D reconstruction is a based on cone-beam filterbackprojection. This method requires the parameters of the X-ray perspective projection, and it also requires the geometry of the C-arm gantry in relation to the patient coordinate system. A calibration procedure is run which involves taking X-ray images from a calibration phantom. The calibration phantom is an acrylic cylinder on which a large number of small steel balls are arranged in a pattern. Figure 2 illustrates this calibration phantom. The arrangement of the spheres is designed such that from any partial view, the geometry of the C-arm and the parameters of the X-ray projection can be determined without ambiguity. Figure 2. X-ray and geometry calibration phantom The images from the calibration phantom are processed by a software application that automatically detects the patterns from the images, and it matches them with the model of the phantom patterns. This 2D to 3D correspondence is used to recover the homogeneous transformation matrix that maps the 3D object to the 2D X-ray projection. This matrix is used in the backprojection process. Figure 3 shows the layout of the calibration software. More detail of this work is described by Navab et al. [1]. Fahrig et al. [2] demonstrated that the Xray perspective transformation and gantry geometry are reproducible. Therefore, the calibration software needs to be run once per installation. Figure 1. X-ray Angio system with C-arm gantry used for data acquisition. Proceedings of the Fourth IEEE Workshop on Applications of Computer Vision (WACV’98) 0-8186-8606-5/98 $17.00 © 1998 IEEE Figure 3. Layout of X-ray geometry calibration software. The X-ray images obtained from the machine have geometry distortion due to; (1) the effect of earth’s magnetic field on the image intensifier and (2) the curvature of the detector surface. A second phantom consisting of a 2D array of small spheres is imaged at every angle of the C-arm. These images are used to compute the warping that is needed to correct for the geometric distortion of the X-ray images. The intensity response of the image intensifier is a nonuniform function of detector surface. This non-uniformity is also corrected for by taking a bright field image (X-ray image of the air). N. Navab et al. [3] give a more complete description of the reconstruction procedure. Figure 4. Fly-Through using interactive volume rendering as the viewing mode. After volume reconstruction we end up with a 3D data set. For the interventionals it is necessary to have visualization tools for effective and quick examination of data. The visualization and post processing takes place on a Siemens post-processing workstation named Virtuoso. Virtuoso is a biomedical visualization platform consisting of several components. Fly-Through is one of the components of Virtuoso that provides an effective means of examining the reconstruction results. It provides several interactive visualization modes all in one framework. The user may examine the 3D data from a global vantage, as well as, from close-up using virtual endoscopy. Figure 4 illustrates FlyThrough software application. Virtual endoscope is an important notion in Fly-Through. It is a tool for navigating around. The user can easily maneuver the endoscope interactively. The endoscope has two important functions: first, it is used to generate endoscopic views from within the organs, secondly, it carries a plane for computing MPR. If the loaded data set contains presegmented organs, then the segmented objects can be converted to models. The models are used by the endoscope for path planning and collision avoidance [4]. Collision avoidance is important, especially for navigating complex structures such as blood vessels. Figure 5 shows the rendering of the arterial model along with the virtual endoscope. A plane orthogonal to the endoscope is attached to the endoscope for the computation of MPRs. This provides a useful tool for examining the vessel cross sections based on raw data. Fly-Through was selected among 45 most innovative technologies by the Discover Magazine [5]. Figure 5. Model of the arterial tree, the virtual endoscope, and the associated plane for computing MPRs. [1] N. Navab et al., Dynamic Geometrical Calibration for 3-D Cerebral Angiography, 1996, SPIE Vol. 2708, pp. 361[2] R. Fahrig et al., Characterization of a C-arm mounted XRII for 3D image reconstruction during interventional neuroradiology, 1996, SPIE Vol. 2708, pp. 351[3] N. Navab et al. 3D Reconstruction From Projection Matrices In A C-Arm Based 3D-Angiography System, MICCAI 98. [4] B. Geiger and R Kikinis", Simulation of endoscopy, Computer Vision, Virtual Reality and Robotics in Medicine, 1995, Vol. 905, pp. 277-281. [5] Discover Magazine, July 1998. Proceedings of the Fourth IEEE Workshop on Applications of Computer Vision (WACV’98) 0-8186-8606-5/98 $17.00 © 1998 IEEE