Search

EP-3948787-B1 - CONE BEAM ARTIFACT CORRECTION FOR GATED IMAGING

EP3948787B1EP 3948787 B1EP3948787 B1EP 3948787B1EP-3948787-B1

Inventors

  • BROWN, Kevin, Martin
  • KOEHLER, THOMAS
  • BONTUS, CLAAS

Dates

Publication Date
20260506
Application Date
20200324

Claims (10)

  1. An imaging system (706), comprising: a reconstructor (716) configured to reconstruct obtained cone beam projection data with a voxel-dependent redundancy weighting such that low frequency components of the cone beam projection data are reconstructed with more redundant data than high frequency components of the cone beam projection data to produce volumetric image data, wherein the reconstructor includes a high pass filter to high pass filter the cone beam projection data and produce high frequency projection data, a low pass filter to low pass filter the cone beam projection data and produce low frequency projection data, a high frequency backprojector (806) configured to backproject the high frequency projection data to produce high frequency volumetric image data, and a low frequency backprojector (808) configured to backproject the low frequency projection data to produce low frequency volumetric image data, wherein the voxel-dependent redundancy weighting includes a high frequency voxel-dependent redundancy weighting and a low frequency voxel-dependent redundancy weighting, wherein the high frequency backprojector is configured to determine a high frequency voxel-dependent redundancy weight for each voxel during backprojection and the high frequency voxel-dependent redundancy weight for at least two voxels is different, and the low frequency backprojector is configured to determine a low frequency voxel-dependent redundancy weight for each voxel during backprojection and the low frequency voxel-dependent redundancy weight for at least two voxels is different, wherein either (i) the high frequency backprojector is configured to determine the high frequency voxel-dependent redundancy weight for each voxel based on a high frequency cardiac weight with a first angular range and an aperture weight for the voxel, and the low frequency backprojector is configured to determine the low frequency voxel-dependent redundancy weight for each voxel based on a low frequency cardiac weight with a second angular range and the aperture weight for the voxel, wherein the first angular range is larger than the second angular range; or (ii) wherein the high frequency backprojector and the low frequency backprojector are configured to determine voxel-dependent redundancy weights based on: CW i ⋅ G w i ∑ j ∈ pp CW j ⋅ G w i , where i represents a current view, CW represents a cardiac weight, G ( w ) represents an aperture weight, and ∑ j ∈ pp () represents a summation of a product of CW and G ( w ) over all pi-partners j of the current view.
  2. The system of claim 1, wherein the high frequency backprojector and the low frequency backprojector are a same backprojector.
  3. The system of any of claims 1 to 2, wherein the reconstructor further includes an adder configured to combine the high frequency volumetric image data and the low frequency volumetric image data to produce the volumetric image data.
  4. A computer-implemented method, comprising: reconstructing obtained cone beam projection data with a voxel-dependent redundancy weighting such that low frequency components of the cone beam projection data are reconstructed with more redundant data than high frequency components of the cone beam projection data to produce volumetric image data, wherein the reconstructing comprises: backprojecting, with a high frequency backprojector, high frequency projection data to produce high frequency volumetric image data; and backprojecting, with a low frequency backprojector, low frequency projection data to produce low frequency volumetric image data, determining a high frequency voxel-dependent redundancy weight for each voxel during backprojection, wherein the high frequency voxel-dependent redundancy weight for at least two voxels is different; and determining a low frequency voxel-dependent redundancy weight for each voxel during backprojection, wherein the low frequency voxel-dependent redundancy weight for at least two voxels is different, wherein the voxel-dependent redundancy weighting includes the high frequency voxel-dependent redundancy weight and the low frequency voxel-dependent redundancy weight, wherein the reconstructing further comprises: determining the high frequency voxel-dependent redundancy weight based on a high frequency cardiac weight and an aperture weight for the voxel, wherein the high frequency voxel-dependent redundancy weight is normalized with rays being two pi apart; and determining the low frequency voxel-dependent redundancy weight based on a low frequency cardiac weight and the aperture weight for the voxel, wherein the low frequency voxel-dependent redundancy weight is normalized across pi partners.
  5. The method of claim 4, wherein the reconstructing further comprises: high pass filtering the projection data to produce the high frequency projection data; and low pass filtering the projection data to produce the low frequency projection data.
  6. The method of any of claims 4 to 5, wherein the reconstructing further comprises: adding the high frequency volumetric image data and the low frequency volumetric image data to produce the volumetric image data.
  7. A computer-readable storage medium storing computer executable instructions which when executed by a processor of a computer cause the processor to perform the method according to claim 4.
  8. The computer-readable storage medium of claim 7, wherein the computer executable instructions further cause the processor to: backproject, with a high frequency backprojector, high frequency projection data to produce high frequency volumetric image data; and backproject, with a low frequency backprojector, low frequency projection data to produce low frequency volumetric image data.
  9. The computer-readable storage medium of claim 8, wherein the computer executable instructions further cause the processor to: high pass filter the projection data to produce the high frequency projection data; and low pass filter the projection data to produce the low frequency projection data.
  10. The computer-readable storage medium of any of claims 8 to 9, wherein the computer executable instructions further cause the processor to: add the high frequency volumetric image data and the low frequency volumetric image data to produce the volumetric image data.

Description

FIELD OF THE INVENTION The following generally relates to imaging and more particularly to cone beam artifact correction for gated imaging and is described with particular application to computed tomography (CT). BACKGROUND OF THE INVENTION Cone beam artifacts can occur in different types of imaging situations in wide coverage computed tomography (CT) systems. One type of cone beam artifact occurs when the X-ray source path meets the criteria for an exact reconstruction (i.e. the source path crosses all planes containing an image point to be reconstructed), but more data is desired to be backprojected than is needed by the exact reconstruction algorithm, e.g., to improve the dose utilization of the scan. This type of cone beam artifact can be corrected, e.g., as described in US 7,027,552 B2, and/or otherwise. The Article by SHECHTER GET AL: "High-Resolution Images of Cone Beam Collimated CT Scans" IEEE TRANSACTIONS ON NUCLEAR SCIENCE, IEEE SERVICE CENTER, NEW YORK, NY, US, vol. 52, no. 1, 1 February 2005, pages 247-255, XP01 1 129945 describes a novel reconstruction algorithm for high-transaxial-rsolution cone beam collimated CT scans, combining a adapted ray-offset technique with a so-called frequency-split technique. The ray-offset technique is adapted by, instead of interleaving parallel re-binned projections from opposite directions during pre-processing (before radial re-binning and filtering), re-binning the parallel projections binned with virtual projections which contain only zeros. The frequency-split technique minimizes zone-beam artifacts which are mainly present in the low-frequency components. A high-frequency image is reconstructed with the ray-offset technique using all redundant data, and a high-pass filter suppresses low-frequency artifacts, while the low-frequency components of the image are reconstructed by using little redundant data only. The Article by HIROYUKI KUDO ET AL: "Exact and approximate algorithms for helical cone- beam CT, Exact and approximate algorithms for helical cone-beam CT", PHYSICS IN MEDICINE AND BIOLOGY, INSTITUTE OF PHYSICS PUBLISHING, BRISTOL GB, XP020023777 describes helical cone-beam CT. The Article by KUDO H ET AL: "New approximate filtered backprojection algorithm for helical cone-beam CT with redundant data", 2003 IEEE NUCLEAR SCIENCE SYMPOSIUM CONFERENCE RECORD, NEW YORK, NY, 19 October 2003, pages 3211-3215, XP01 0742770 describes helical cone-beam CT. A different situation occurs in scans where the source path does not cross all planes containing an object point to be reconstructed and an exact reconstruction is not possible. Two such cases are illustrated in FIGS. 1 and 2, respectively for an axial scan and a gated helical scan. In FIG. 1, source locations 102 (the thicker open arc) on a circular path 104 (the thinner closed loop) of an X-ray source 106 that correspond to a desired cardiac phase to reconstruct do not cross an imaging plane 108 (i.e. all object points in the plane cannot be reconstructed exactly). In FIG. 2, source locations 202 (the thicker open arc) on a helical path 204 (the thinner helix) of an X-ray source 206 and corresponding to a desired cardiac phase to reconstruct do not cross an imaging plane 208. A common gating approach used to reconstruct higher temporal resolution volumetric image data is Parker-Weighting, which is shown in FIG. 3. In FIG. 3, a first axis 302 represents detector position and a second axis 304 represents source angle. The cone-angle corresponding to the detector row is not shown because all rows are typically treated the same way except for a weighting with the cosine of the cone-angle. With this approach, only a subset 306 (e.g., 220-degrees, or 180-degrees 308 plus a beam angle 310, of parallel views) of acquired projections 312 (e.g., for a 232-degree scan) or 314 (e.g., for a 360-degree scan) are used to reconstruct the image. A weight is applied to the subset 306 before backprojection and, thus, the weight must apply to all voxels in a same manner during backprojection. Such a weight is referred to as a voxel-independent (i.e. global) redundancy weight. Unfortunately, such a voxel-independent (i.e. global) redundancy weight is not well-suited for three-dimensional (3-D) reconstructions from a wide cone beam acquisition. An example of this is discussed in connection with FIG. 4. FIG. 4 shows a source 402 at a first angular position 404 and a cone beam (z-axis only) 406 for a z-axis coverage 408 of a detector. FIG. 4 also shows, overlaid therewith, the source 402 at a second angular position 410, which is 180-degrees opposite the first angular position 404, and a cone beam (z-axis only) 412 for the z-axis coverage 408 of the detector. A ray 414 from the first angular position 404 is within the cone beam 406 and thus is detected by the detector and traverses an image plane 416 at a position 418. A ray 420 from the second angular position 410 is outside of the cone beam 412 and thus is not detected by the d