3D region of interest X-ray CT for geometric magnification from multiresolution acquisitions

We present an algorithm which allows us to reconstruct a region-of-interest (ROI) from transaxial truncated cone-beam projections. While sometimes it is possible to reconstruct good approximations from projections passing only through a specific localized region, in other cases the reconstruction of the original density function shows severe artifacts, especially when the contaminating part outside the ROI is made of unknown dense structures. The main goal of this work is the improvement of resolution of the reconstruction by placing an object closer to the X-ray source so only the ROI is within the field of view, allowing a substantial geometric magnification. We will use the well-known fact that a rough estimate of the global properties of the Radon transform is sufficient to obtain a reliable reconstruction of a local region. The successful and widely used Feldkamp-algorithm (FDK) is applied on the combination of global projections acquired at a low resolution with transaxial truncated high-resolution acquisitions. Moreover, we also investigate the effect of interpolation between this low-resolution acquisitions, in order to reduce the X-ray radiation doses.