True 3D cone-beam imaging method and apparatus

• US 6,865,246 B2
• Filed: 03/26/2003
• Issued: 03/08/2005
• Est. Priority Date: 09/26/2001
• Status: Expired due to Fees

• Current Assignee: Mit Technology Licensing Office
• Original Assignee: Massachusetts Institute of Technology
• Inventors: Yang, Xiaochun
• Primary Examiner(s): BRUCE, DAVID VERNON
• Application Number:
  US10/400,344
• Publication Number:
  US 20030202637A1
• Time in Patent Office: 713 Days
• Field of Search: 378/4, 378/8, 378/15, 378/19, 378/901
• US Class Current: 378/4
• CPC Class Codes: A61B 6/027 characterised by the use of a particular data acquisition trajectory, e.g. helical or spiral
• CPC Class Codes: A61B 6/032 Transmission computed tomography [CT]
• CPC Class Codes: A61B 6/4085 Cone-beams
• 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 2211/421 Filtered back projection [FBP]
• CPC Class Codes: Y10S 378/901 Computer tomography program or processor

Abstract

A three-dimensional image of an object scanned with a plurality of cone-beam projections from a number of source positions is reconstructed using a method wherein intermediate transform functions are obtained from two-dimensional images of radiation attenuation in the scanned object. The intermediate transform functions are then filtered over a set of parallel planes using a moving-frame technique. The second-order radial derivative of the Radon transform can then be backprojected to generate an intermediate, locally-reconstructed, three-dimensional image. After repetition of this process, the plurality of intermediate, locally reconstructed, three-dimensional images are summed to obtain an ultimate, reconstructed, three-dimensional image of the object. In particular embodiments, the source and detector are displaced along helical paths and radiation scans of the object are taken at multiple positions along the paths.

Claims

1. A method for generating a three-dimensional image of a scanned object from a plurality of cone-beam projections passed through the object and attenuated thereby, the method comprising:

a) positioning a source at a position on a predetermined scan path;

b) passing a projection of cone-beam radiation comprising a plurality of projection rays from the source through an object, the cone-beam projection being attenuated by partial absorption in the object;

c) detecting radiation intensity of the attenuated cone-beam projection on an area detector;

d) obtaining a two-dimensional attenuation image of the cone-beam projection from the detected radiation intensity;

e) obtaining an intermediate transform function from the two-dimensional attenuation image on a set of planes passing through the source;

then

f) repeatedly;

i) displacing the source to another position on the predetermined scan path and then

ii) repeating steps (b)-(e);

g) filtering the intermediate transform functions acquired from consecutive attenuation images at two or more source positions using a moving-frame technique to obtain the second-order radial derivative of the Radon transform;

h) backprojecting the second-order radial derivative of the Radon transform in a two-dimensional space along each projection ray passing through a source position among the source positions referenced in step (g) to generate an intermediate, locally reconstructed, three-dimensional image with constant values assigned along each projection ray;

i) at least once, repeating steps (f)-(h); and

j) summing the plurality of intermediate, locally reconstructed, three-dimensional images obtained for the plurality of cone-beam projections to obtain an ultimate, reconstructed, three-dimensional image of the object.

2. The method of claim 1, wherein the predetermined scan path is everywhere differentiable with exception allowed at a finite number of points.

3. The method of claim 1, wherein the intermediate transform function is the first-order radial derivative of the Radon transform.

4. The method of claim 3, wherein each two-dimensional attenuation image is obtained as a discrete two-dimensional data set with an area detector comprising a plurality of detector elements, and wherein the step of obtaining the first-order radial derivative of the Radon transform includes:

i) interpolating data points lying between grid nodes along a set of virtual image lines;

ii) calculating a weighted sum of attenuation along each virtual image line with weights determined by the cosine of the angle between (a) a projection ray associated to an image point and (b) a central ray, which is the shortest line from the source to the virtual image line;

iii) differentiating the weighted sum over a set of parallel virtual image lines to obtain the first-order radial derivative of the Radon transform on a set of corresponding planes passing through the source; and

iv) storing the resulting first-order radial derivative of the Radon transform in memory for use in a subsequent phase of reconstruction.

5. The method of the claim 4, wherein, if the area detector does not coincide with the standard image plane, the method further comprises:

a) mapping the coordinates of the detector elements to the coordinates of a set of points on a standard image plane under a projective coordinate transform; and

b) interpolating the two-dimensional data set to obtain attenuation values on a regular grid on the standard image plane.

6. The method of claim 4, wherein the virtual image lines pass through a set of grid nodes on a standard image plane, wherein the same number of virtual image lines of different slopes passing through each grid node, and wherein the collection of slopes of the intersecting virtual image lines is the same at each grid node.

7. The method of claim 4, wherein the regular grid has square or rectangular elements.

8. The method of claim 1, wherein the two-dimensional backprojection along a projection ray is an integral over all planes passing by that ray, wherein the unit normal vectors of these planes are confined to a great circle on the unit sphere.

9. The method of claim 1, wherein a plurality of cone-beam projections are successively passed through the object, and wherein the step of determining the second-order radial derivative of the Radon transform includes:

i) determining the global coordinates, represented by the unit normal and radial distance, of a set of corresponding planes pass though a set of virtual image lines on the standard image plane and through the current focal point;

ii) obtaining the first-order radial derivative of the Radon transform on a set of planes for the next one or more cone-beam projections;

iii) calculating the local parameters of intersection image lines between standard image planes in successive cone-beam projections and corresponding planes in successive cone-beam projections parallel to the set of planes passing through the current focal point via a coordinate transformation determined by a moving-frame technique;

iv) interpolating the first-order radial derivative of the Radon transform for the planes obtained in part (iii) from the first-order radial derivative of the Radon transform obtained in part (ii); and

v) obtaining the second-order radial derivative of the Radon transform by differentiating, which can be approximated by a finite difference, the first-order radial derivative of the Radon transform over the parallel planes from consecutive cone-beam projections.

10. The method of claim 1, wherein the step of backprojecting the second-order radial derivative includes:

for each projection ray associated with a node on the image plane, calculating a weighted sum of the second-order radial derivative of the Radon transform over a set of planes passing through the projection ray with weights determined by (a) the number of intersections of each plane with the scan path, (b) the absolute value of a dot product between a tangent vector of the scan path at the current source position and a unit normal vector of the plane, and (c) the angular interval between two adjacent planes;

multiplying the weighted sum by an incremental value of the source orbit parameter, λ

, from the current source position to a next source position;

assigning the resulting value to points lying on the projection ray; and

resampling and interpolating the projection rays so that each node on a three-dimensional Cartesian grid representing the object is positioned on one of the rays and is assigned a value to produce a three-dimensional, locally reconstructed image.

11. The method of claim 1, wherein the radiation is X-ray radiation.

12. The method of claim 1, wherein the detector is aligned according to at least one of the following:

a) the detector is placed orthogonal to a line connecting the source to the global-coordinate origin, and an axis of the detector is aligned with the projection of the tangent of the scan path at a current source position onto the detector; and

b) the detector is place with a first axis parallel with a rotational axis of the scan path and a second axis parallel with a projection of the tangent of the scan path at a current source position onto a plane perpendicular to the rotational axis of the scan path.

13. The method of claim 1, wherein the moving frame technique involves the generation of a set of orthonormal bases with the origin of each orthonormal basis positioned at the source such that each time the source is repositioned and the two-dimensional attenuation image is obtained, the orthonormal basis has a new origin, wherein the moving frame technique enables locating a set of parallel planes from consecutive cone-beam projections and enables the second-order radial derivative of the Radon transforms to be evaluated by filtering over the set of parallel planes from consecutive cone-beam projections.

14. The method of claim 13, wherein the moving frame technique enables location of the planes in a subsequent cone-beam projection parallel to the set of planes in the current projection via a set of coordinate transforms, the method further comprising:

i) for each selected line on a current standard image plane, finding global coordinates, which can be expressed as l and β

, in the Radon space of a corresponding plane passing through a focal point of the cone-beam radiation and intersecting the standard image plane at the selected line;

ii) finding global coordinates, which can be expressed as l and β

, of planes passing through the focal point of the subsequent cone-beam projection and parallel to the planes obtained in part (i), wherein two parallel planes have the same unit normal but different radial distances from the global origin; and

iii) finding local coordinates, which can be expressed as polar coordinate, s and ø

, on a standard image plane, of intersection lines between the planes obtained in part (ii) and a standard image plane of the subsequent cone-beam projection.

15. The method of claim 1, wherein the source and the detector are attached to a rigid arm that is mounted to a rotary motor that axially rotates the rigid arm.

16. The method of claim 15, wherein the rigid arm includes one or more slides to which the source and detector are attached, the slides displacing the source and detector along an axis parallel to the axis of rotation of the rigid arm.

17. The method of claim 15, wherein the object is displaced along the axis of rotation of the rigid arm.

18. The method of claim 2, wherein the detector is not large enough to detect the entirety of a projection of radiation passing from the source through the entirety of the object.

19. The method of claim 18, wherein the predetermined scan path is a helical scan path;

and the detected radiation is masked using a Tam window to eliminate redundancy, wherein backprojecting the second-order radial derivative further comprises an extra step of adding a contribution of the second-order radial derivative on processed planes to reconstructions at other source points that pass through the same plane, the extra step including;

for every plane where the second order radial derivative of the Radon transform is evaluated, finding a set of intersections between the plane and the helical scan path, which can be denoted by Φ

(λ

⁽ⁱ⁾) (i=2,3, . . . , M(Φ

(λ

⁽¹⁾)·

β

, β

)), where Φ

(λ

) is the parameterized scan path, β

is the unit normal of the said plane and M(Φ

(λ

⁽¹⁾)·

β

, β

) is the multiplicity function which equals the number of the intersections of the plane and the helical scan path;

finding the source points nearest to Φ

(λ

⁽ⁱ⁾) (i=2,3, . . . , M(Φ

(λ

⁽¹⁾)·

β

, β

)), which can be denoted by Φ

(λ

^(i)′

) (i=2,3, . . . , M(Φ

(λ

⁽¹⁾)·

β

, β

)), on the sampled scan path; and

adding a contribution of the evaluated second-order radial derivative on the plane, which can be denoted by R″

ƒ

(Φ

(λ

⁽¹⁾)·

β

, β

), to each projection line passing through Φ

(λ

^(i)′

) i=2,3, . . . , M(Φ

(λ

⁽¹⁾)·

β

, β

), and on the plane through Φ

(λ

^(i)′

) and having unit normal β

, wherein the amount of contribution can be approximated by R″

ƒ

(Φ

(λ

⁽¹⁾)·

β

, β

)|Φ

(λ

⁽ⁱ⁾)·

β

⁽ⁱ⁾|Δ

θ

⁽ⁱ⁾, where β

⁽ⁱ⁾ is the normal vector of a plane (a) nearest to the plane through Φ

(λ

^(i)′

) and having unit normal β

;

(b) being evaluated in the local frame originating at Φ

(λ

^(i)′

), where Δ

θ

⁽ⁱ⁾ is the angular interval between neighboring planes evaluated in the local frame originating at Φ

(λ

^(i)′

) and passing the projection line.

20. A cone-beam tomography apparatus comprising a radiation source, a radiation detector, a support for an object to be scanned by radiation from the radiation source, a computer-readable storage medium storing computer-executable software for generating a reconstruction of cone-beam radiation attenuation in an object, the software comprising:

code for obtaining an intermediate transform function on a set of planes from a signal representing the amount of radiation transmitted through the object;

code for determining the second-order radial derivative of the Radon transform by filtering the intermediate transform function over a set of parallel planes using a moving-frame technique;

code for backprojecting the second-order radial derivative of the Radon transform to generate an intermediate, locally reconstructed, three-dimensional image with constant values assigned along each projection ray;

code for summing the plurality of intermediate, locally reconstructed, three-dimensional images obtained for the cone-beam projections to obtain an ultimate, reconstructed, three-dimensional image of the object; and

code for displacing the source and detector relative to the support in a predetermined scan path for radiation transmitted from the source, through an object positioned by the support, and to the detector.

21. The apparatus of claim 20, wherein the predetermined scan path is everywhere differentiable with exception allowed at a finite number of points.

22. The apparatus of claim 20, further comprising a rigid arm to which the source and the detector are attached.

23. The apparatus of claim 22, wherein the rigid arm is mounted to a rotary motor to allow the rotary motor to axially rotate the rigid arm.

24. The method of claim 23, wherein the rigid arm includes one or more slides to which the source and detector are attached, the slides allowing displacement of the source and detector along an axis parallel to the axis of rotation of the rigid arm.