Projector/backprojector with slice-to-slice blurring for efficient 3D scatter modeling

• US 6,381,349 B1
• Filed: 11/10/1998
• Issued: 04/30/2002
• Est. Priority Date: 11/12/1997
• Status: Expired due to Fees

• Current Assignee: University of Utah (State of Utah)
• Original Assignee: University of Utah (State of Utah)
• Inventors: Bai, Chuanyong, Zeng, Gengsheng Lawrence, Gullberg, Grant T.
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US09/189,686
• Time in Patent Office: 1,267 Days
• Field of Search: 382/128, 382/129, 382/130, 382/131, 382/132, 250/363.04, 358/75, 342/25, 364/413.04, 324/307, 378/4, 378/98.3, 378/98.4, 378/95, 378/87
• US Class Current: 382/128
• CPC Class Codes: G06T 11/005 Specific pre-processing for tomographic reconstruction, e.g. calibration, source positioning, rebinning, scatter correction, retrospective gating
• CPC Class Codes: G06T 2200/04 involving 3D image data
• CPC Class Codes: G06T 2207/10108 Single photon emission computed tomography [SPECT]
• CPC Class Codes: G06T 2207/30004 Biomedical image processing
• CPC Class Codes: G06T 5/20 using local operators
• CPC Class Codes: G06T 5/70 Denoising; Smoothing

Abstract

A method of modeling 3D first-order scatter, non-uniform attenuation, and 3D system geometric point response in an ML-EM algorithm to reconstruction SPECT data is provided. It includes performing an initial slice-to-slice blurring operation (112) on a volume of estimated emission source data. The volume of estimated emission source data is represented by a 3D array of voxels. A voxel-by-voxel multiplying (114) of the results from the initial slice-to-slice blurring operation (112) by a volume of attenuation coefficients yields a volume of effective scatter source data (116). The volume of effective scatter source data (116) is voxel-by-voxel added (118) to the volume of estimated emission source data to produce a volume of combined estimated emission and scatter source data. Finally, a secondary slice-to-slice blurring operation (120) is performed on the volume of combined estimated emission and scatter source data.

Claims

1. A method of modeling 3D first-order scatter, non-uniform attenuation, and 3D system geometric point response in an ML-EM algorithm to reconstruction SPECT data comprising:

(a) performing an initial slice-to-slice blurring on a volume of estimated emission source data, said volume of estimated emission source data being represented by a 3D array of voxels;

(b) voxel-by-voxel multiplying step (a) results by a volume of attenuation coefficients to produce a volume of effective scatter source data;

(c) voxel-by-voxel adding the volume of effective scatter source data to the volume of estimated emission source data to produce a volume of combined estimated emission and scatter source data; and

,

(d) performing a secondary slice-to-slice blurring on the volume of combined estimated emission and scatter source data.

2. A method of medical imaging comprising:

(a) performing an initial slice-to-slice blurring on a volume of estimated emission source data by successively convolving parallel slices of the volume of estimated emission source data with respective blurring kernels, and successively adding previously convolved neighboring slices to their immediately following slices prior to that following slices'"'"' convolution;

(b) voxel-by-voxel multiplying step (a) results by a volume of attenuation coefficients to produce a volume of effective scatter source data;

(c) voxel-by-voxel adding the volume of effective scatter source data to the volume of estimated emission source data to produce a volume of combined estimated emission and scatter source data; and

,

(d) performing a secondary slice-to-slice blurring on the volume of combined estimated emission and scatter source data.

3. The method according to claim 2, wherein after addition of the previously convolved neighboring slice and before its own convolution, slices are point-by-point multiplied by a 2D array of attenuation factors.

4. The method according to claim 3, wherein the attenuation factors are exponential functions whose exponent is a negative of a linear attenuation coefficient times a distance between respective voxels.

5. The method according to claim 2, wherein each convolution is implemented as two orthogonal 1D convolutions.

6. The method according to claim 2, wherein the convolutions approximate a Gaussian scattering probability which is a function of a scattering angle.

7. The method according to claim 6, wherein the blurring kernels are estimated from a least-squares fit comparison of calculated results to projections from known point sources taken one point source at a time using Monte Carlo simulations.

8. A method of medical imaging comprising:

(a) performing an initial slice-to-slice blurring on a volume of estimated emission source data;

(b) voxel-by-voxel multiplying step (a) results by a volume of attenuation coefficients to produce a volume of effective scatter source data;

(c) voxel-by-voxel adding the volume of effective scatter source data to the volume of estimated emission source data to produce a volume of combined estimated emission and scatter source data; and

,

(d) performing a secondary slice-to-slice blurring on the volume of combined estimated emission and scatter source data by successively convolving parallel slices of the volume of combined estimated emission and scatter source data with respective blurring kernels, and successively adding previously convolved neighboring slices to their immediately following slices prior to that following slices'"'"' convolution.

9. The method according to claim 8, wherein after addition of the previously convolved neighboring slice and before its own convolution, slices are multiplied by a 2D array of attenuation factors in a point-by-point manner.

10. The method according to claim 9, wherein the attenuation factors are exponential functions whose exponent is a negative of a linear attenuation coefficient times a distance between respective voxels.

11. The method according to claim 8, wherein the blurring kernels are cross shaped.

12. The method according to claim 8, wherein the convolutions approximate a system geometric point response function, said system geometric point response function being dependent on physical characteristics of a collimator used during collection of the SPECT data.

13. A projector/backprojector for use in an image processor employing an EM reconstruction algorithm comprising:

a first convolver which successively convolves parallel slices of a volume of image data and adds them to neighboring slices prior to the neighboring slices being convolved, said first convolver employing convolution kernels determined from a Compton scattering function which is a function of a scattering angle;

a multiplication processor which voxel-by-voxel multiplies a volume of image data from the first convolver by a volume of attenuation coefficients; and

,

a second convolver which successively convolves parallel slices of a volume of image data from the multiplication processor and adds them to neighboring slices prior to the neighboring slices being convolved, said second convolver employing convolution kernels determined from a system geometric point response function which depends on physical characteristics of a collimator used during collection of measured data.

14. The projector/backprojector according to claim 13, wherein the first and second convolvers point-by-point multiply a corresponding 2D array of attenuation factors with each of the slices prior to its convolution and after addition of its prior convolved neighboring slice.

15. The projector/backprojector according to claim 14, wherein the attenuation factors represent exponential functions whose exponent is a negative of a negative of a linear attenuation coefficient times a distance between respective voxels.

16. A projector/backprojector for use in an image processor employing, said projector/backprojector comprising:

first convolving means for performing a first slice-to-slice blurring operation on a volume of estimated emission source data;

a multiplication processor which voxel-by-voxel multiplies an output from the first convolver by a volume of attenuation coefficients to produce a volume of effective scatter source data;

an adder which adds the volume of effective scatter source data to the volume of estimated emission source data; and

,

a second convolving means for performing a second slice-to-slice blurring operation on a volume of data output from the adder.