TECHNIQUE TO CORRECT POSITRON EMISSION TOMOGRAPHY AXIAL MISPOSITIONING
Publication Date: 2015-Feb-25
The IP.com Prior Art Database
The present invention proposes a technique to correct positron emission tomography (PET) axial mispositioning. The technique includes comparing image and physical positions of PET axial sinogram bins (z and θ dimensions). Such comparison allows repositioning counts in the sinogram through an interpolation technique that corrects for the mis-positioning. The axial repositioning is applied in sinogram space by use of weights which are precalculated for each configuration. As the axial geometry and slice thickness does not ever change from scan to scan, the weights are required to be calculated once.
The present invention relates generally to positron emission tomography (PET) and more particularly to a technique for correcting PET axial mispositioning.
In a PET detector system detectors in axial direction are not uniformly spaced. Usually, gaps are present between adjacent detector blocks. The gaps provide shifts in positions of reconstructed images unless accounted for by a reconstruction algorithm. Due to slow process of the reconstruction algorithm, direct compensation by reconstruction algorithm is computationally very expensive. As a result, most reconstruction techniques assume that image slices are completely adjacent to each other without any gaps. Consequently, positioning errors exist in apparent image and position of image slices does not line up exactly with their physical locations. Figure 1 depicts misalignment between physical position of the image slices and the assumed location in image space in the axial direction.
As illustrated in Figure 1, physical position of the image slices is depicted in yellow color. Assumed location in image space in axial direction is depicted in blue color. Physical location of detector crystals is depicted in red color.
The positioning error depends upon relative sizes of the detector crystals and the gaps. However, it is possible to have several millimeters of positioning. Figure 2 depicts a graph between slice position shifts and slice thickness. The graph infers that mispositioning error is as high as a few millimeters.
Further, PET raw data is stored in a four-dimensional histogram array known as a sinogram. Each element of the array represents a line-of-response (LOR) that connects two crystals. Four dimensions are required to uniquely specify the LOR. The four dimensions include r, φ, z and θ, where, (r, φ) describes line in trans-axial plane and (z, θ) describes line in orthogonal axial direction.
Due to computational overhead of reconstructing with exact position of the crystals, conventional techniques reconstruct assuming no gaps between image slices. Each image slice is slightly enlarged so as to fill physical axial field-of-view.
A conventional technique includes adding pseudo-slices to PET sinograms to correct for slice position errors arising from axial gaps. Although the technique is geometrically intuitive, however, the technique poses certain implementation issues as the number of slices per segment does not follow the usual pattern of a difference by four between segments.
Another conventional technique includes utilization of pseudo-crystals to correct for axial mispositioning of PET image slices. However, the technique requires certain changes in software architecture.
It would be desirable to have an efficient technique to correct PET axial mispositioning.
BRIEF DESCRIPTION OF THE FIGURES
Figure 1 depicts misalignment between physical position of...