Search

In this paper, three exact formulae are derived for cone-beam reconstruction with source positions on a curve or set of curves. For reconstruction at a single point, these formulae all operate by applying a filtration step followed by a backprojection step to cone-beam data. The filtering is performed along a 1D curve which is defined as the intersection of the detector surface with a filtering plane. Two of these formulae allow a flexibility in the choice of the filtering direction. In some cases, this flexibility allows the efficiency of volume reconstruction to be improved. Alternatively, the flexibility can be used to reduce the detector size necessary to avoid truncation artefacts in the reconstruction or to change the noise properties of the reconstruction.

We derive unified inversion formulae for the cone beam transform similar to the Radon transform. Reinterpreting Grangeat's formula we find a relation between the Radon transform of the gradient of the searched-for function and a quantity computable from cone beam data. This gives a uniqueness result for the cone beam transform of compactly supported functions under much weaker assumptions than the Tuy–Kirillov condition. Furthermore this relation leads to an exact formula for the direct calculation of derivatives of the density distribution; but here, similar to the classical Radon transform, complete Radon data are needed, hence the Tuy–Kirillov condition has to be imposed. Numerical experiments reported in Hahn B N et al (2013 Meas. Sci. Technol. 24 125601) indicate that these calculations are less corrupted by beam-hardening noise. Finally, we present flat detector versions for these results, which are mathematically less attractive but important for applications.

The mechanical motion of the gantry in conventional cone beam CT scanners restricts the speed of data acquisition in applications with near real time requirements. A possible resolution of this problem is to replace the moving source detector assembly with static parts that are electronically activated. An example of such a system is the Rapiscan Systems RTT80 real time tomography scanner, with a static ring of sources and axially offset static cylinder of detectors. A consequence of such a design is asymmetrical axial truncation of the cone beam projections resulting, in the sense of integral geometry, in severely incomplete data. In particular we collect data only in a fraction of the Tam–Danielsson window, hence the standard cone beam reconstruction techniques do not apply. In this work we propose a family of multi-sheet surface rebinning methods for reconstruction from such truncated projections. The proposed methods combine analytical and numerical ideas utilizing linearity of the ray transform to reconstruct data on multi-sheet surfaces, from which the volumetric image is obtained through deconvolution. In this first paper in the series, we discuss the rebinning to multi-sheet surfaces. In particular we concentrate on the underlying transforms on multi-sheet surfaces and their approximation with data collected by offset multi-source scanning geometries like the RTT. The optimal multi-sheet surface and the corresponding rebinning function are found as a solution of a variational problem. In the case of the quadratic objective, the variational problem for the optimal rebinning pair can be solved by a globally convergent iteration. Examples of optimal rebinning pairs are computed for different trajectories. We formulate the axial deconvolution problem for the recovery of the volumetric image from the reconstructions on multi-sheet surfaces. Efficient and stable solution of the deconvolution problem is the subject of the second paper in this series (Betcke and Lionheart 2013 Inverse Problems 29 115004).

In airport baggage scanning it is desirable to have a system that can scan baggage moving at standard conveyor belt speeds. One way to achieve this is to use multiple electronically switched sources rather than a single source on a mechanically rotated gantry. In such a system placing the detectors opposite the sources would obstruct the beam, so they have to be offset (hence offset multi-source geometry). This results in asymmetrical axial truncation of the cone beam projections. As such projections do not constitute complete data in the sense of integral geometry, the standard cone beam reconstruction algorithms do not apply. In this series of papers we introduce a new family of rebinning methods for reconstruction from axially asymmetrically truncated cone beam projections. In the first paper we discussed the approximation of the data on the multi-sheet surface with the truncated projection data obtained from offset multi-source geometries. In this second paper we focus on the recovery of the volumetric image from the reconstruction of data rebinned to multi-sheet surfaces. Multi-sheet rebinning effects an implicit relation between the fan beam transforms on the individual sheets and the rebinned data. This relation in conjunction with the linearity of the ray transform allows us to formulate the deconvolution problem for the recovery of the volume from a stack of reconstructed images on multi-sheet surfaces. We discuss the errors in the right-hand side of the deconvolution problem (reconstruction on multi-sheet surfaces) resulting from rebinning approximation. We introduce convolution matrix models based on the distribution of the distances of the rays from the multi-sheet surface, which considerably improve the data model fit and in turn lead to a superior solution. Multiple strategies for solution of the deconvolution problem are discussed and an efficient and robust implementation is presented, which makes the method capable of real time reconstruction. We conclude with some reconstruction results from simulated as well as real data collected with a Rapiscan RTT80 scanner.