Method and apparatus for reconstruction of 3D image volumes from projection images

• US 7,693,318 B1
• Filed: 10/31/2005
• Issued: 04/06/2010
• Est. Priority Date: 01/12/2004
• Status: Active Grant

• Current Assignee: PME IP Pty., Ltd.
• Original Assignee: PME IP Australia Pty Ltd. (Pro Medicus Limited)
• Inventors: Seebass, Martin, Stalling, Detlev, Kubis, Ralf, Westerhoff, Malte
• Primary Examiner(s): Seth; Manav
• Application Number:
  US11/263,568
• Time in Patent Office: 1,618 Days
• Field of Search: 382/128, 382/131, 382/132, 382/285, 382/293, 378/4, 378/10, 378/21, 378/23, 128/922
• US Class Current: 382/128
• CPC Class Codes: G06T 11/006 Inverse problem, transformation from projection-space into object-space, e.g. transform methods, back-projection, algebraic methods
• CPC Class Codes: G06T 2211/428 Real-time
• CPC Class Codes: Y10S 128/922 including image analysis

Abstract

The invention provides improvements in reconstructive imaging of the type in which a volume is reconstructed from a series of measured projection images (or other two-dimensional representations) by utilizing the capabilities of graphics processing units (GPUs). In one aspect, the invention configures a GPU to reconstruct a volume by initializing an estimated density distribution of that volume to arbitrary values in a three-dimensional voxel-based matrix and, then, determining the actual density distribution iteratively by, for each of the measured projections, (a) forward-projecting the estimated volume computationally and comparing the forward-projection with the measured projection, (b) generating a correction term for each pixel in the forward-projection based on that comparison, and (c) back-projecting the correction term for each pixel in the forward-projection onto all voxels of the volume that were mapped into that pixel in the forward-projection.

Claims

1. An apparatus for computed tomography of the type that reconstructs density distribution of a volume from a plurality of measured projection images, comprising

a graphics processing unit (GPU) including a plurality of programmable vertex shaders that are coupled to a plurality of programmable pixel shaders,

the vertex shaders and the pixel shaders being adapted to generate, for at least one measured projection image, a respective forward-projection image of an estimated density distribution assigned to the volume and to generate one or more correction terms based on comparison of one or more pixels in the forward-projection image with respective pixels in that measured projection image.

2. The apparatus of claim 1, the further wherein the volume is represented as a plurality of slices.

10. The apparatus of claim 1, wherein the GPU classifies the measured projection images into two groups characterized by substantially orthogonal projection directions.

11. The apparatus of claim 10, wherein the GPU processes the measured projection images in one group prior to processing the measured projection images in the other group.

12. The apparatus of claim 11, wherein the GPU processes the measured projections in at least one of the groups according to a quasi-random sequence.

3. An apparatus for computed tomography of the type that reconstructs density distribution of a volume from a plurality of measured projection images, comprising

a graphics processing unit (GPU) including a plurality of programmable vertex shaders that are coupled to a plurality of programmable pixel shaders,

the vertex shaders and the pixel shaders being adapted to generate for at least one measured projection image a respective forward-projection image of an estimated density distribution assigned to the volume and to generate one or more correction terms based on comparison of one or more pixels in the forward-projection image with respective pixels in that measured projection image;

wherein the volume is represented as a plurality of slices, and

wherein one or more of the vertex shaders are configured to map to each pixel in the forward-projection image a plurality of voxels that lie substantially at vertices of one or more selected polygonal portions of the slices.

4. The apparatus of claim 3, wherein the one or more vertex shaders treat the forward-projection image as a “

window” and

each of the polygonal portions as a “

model”

being projected onto the window.

5. The apparatus of claim 3, wherein one or more of the pixel shaders compute an intensity for each pixel in the forward-projection image as a summation of the projections of the voxels mapped to that pixel.

6. The apparatus of claim 5, wherein at least one of the pixel shaders computes multiple forward-projection steps along a ray, which extends through the volume to a pixel in the forward-projection image, prior to computing corresponding steps for another ray.

7. The apparatus of claim 4, wherein at least one of the pixel shaders generates correction terms for one or more pixels in the forward-projection image based on comparison of the intensities of those pixels with intensities of the respective pixels in the measured projection image.

8. The apparatus of claim 4, wherein at least one of pixel shaders back-projects the correction term corresponding to each pixel in the forward-projection image to voxels in the volume that are mapped to that pixel.

9. The apparatus of claim 8, wherein the one or more pixel shaders generate an updated density distribution for the volume as a weighted sum of the estimated density distribution at each voxel and the back-projected correction term corresponding to that voxel.

13. The apparatus of claim 3, wherein the reconstruction volume comprises an intersection of cone beams corresponding to the measured projection images.

14. The apparatus of claim 8, wherein the at least one pixel shader back-projecting the correction term utilizes a windowing function to weight the correction term along one or more back-projection rays.

15. The apparatus of claim 14, wherein one of the pixel shaders computes the weights applied to the back-projected correction term.

16. The apparatus of claim 15, wherein the weights are computed via lookup in a 1-D texture.