Experimental investigation of arbitrary-orientation cone-beam X-ray tomography

X-ray Computed Tomography is a technique to reconstruct an image of trans-axial slab of the object from a series of x-ray radiographs taken at a prior-known angle. Sequences of x-ray radiographs are served as two-dimensional projection data for a 3D tomography. The most popular Feldkamp Algorithm which is based on Filtered Backprojection (FBP) approaches has shown to perform well for 3D reconstruction. In the case of limited view, however, Feldkamp Algorithm suffers from star artifact. In these scenarios, an algebraic reconstruction technique such as the Simultaneous Algebraic Reconstruction Technique (SART) is engaged for reconstructing tomograms. Conventional x-ray computed tomography was implemented on a c-arm x-ray apparatus where the x-ray source and detector is capable of rotating to capture radiograph at any specific angle. The implementation of conebeam - geometry reconstruction algorithm, however, requires that the center location of the detector is accurately identified. Any slightly-missed alignment of the x-ray source or the detector could result in the error of the position of the center and hence the error in reconstructed image. Consequently, x-ray radiography tomography is normally implemented on a c-arm x-ray apparatus where the correct orientation of x-ray tube with respect to x-ray detector is achievable. The aim of this paper is to implement the x-ray tomography on a non c-arm x-ray apparatus where the x-ray source can be in any orientation with respect to x-ray detector. To determine the orientation, we take the radiograph of the reference transparent object, say the plastic box, of which the coordinate of the landmark, say the corner point, is known. The shadowgram of the box is analyzed to extract the coordinate of landmark image and to determine the orientation matrix using classical direct linear transform method (DLT). Once the orientation is known, modified conebeam tomography is performed to derived 3D reconstruction volumetric data. The experimental results demonstrated the potential of such method.